Mathematical Methods
in Quantum Mechanics
With Applications to Schr¨odinger Operators
Gerald Teschl
Gerald Teschl
Fakult¨at f¨ ur Mathematik
Nordbergstraße 15
Universit¨at Wien
1090 Wien, Austria
Email: Gerald.Teschl@univie.ac.at
URL: http://www.mat.univie.ac.at/˜gerald/
2000 Mathematics subject classiﬁcation. 8101, 81Qxx, 4601
Abstract. This manuscript provides a selfcontained introduction to math
ematical methods in quantum mechanics (spectral theory) with applications
to Schr¨odinger operators. The ﬁrst part covers mathematical foundations
of quantum mechanics from selfadjointness, the spectral theorem, quantum
dynamics (including Stone’s and the RAGE theorem) to perturbation theory
for selfadjoint operators.
The second part starts with a detailed study of the free Schr¨odinger op
erator respectively position, momentum and angular momentum operators.
Then we develop WeylTitchmarsh theory for SturmLiouville operators and
apply it to spherically symmetric problems, in particular to the hydrogen
atom. Next we investigate selfadjointness of atomic Schr¨odinger operators
and their essential spectrum, in particular the HVZ theorem. Finally we
have a look at scattering theory and prove asymptotic completeness in the
short range case.
Keywords and phrases. Schr¨odinger operators, quantum mechanics, un
bounded operators, spectral theory.
Typeset by /
´
oL
A
T
E
X and Makeindex.
Version: April 19, 2006
Copyright c ( 19992005 by Gerald Teschl
Contents
Preface vii
Part 0. Preliminaries
Chapter 0. A ﬁrst look at Banach and Hilbert spaces 3
'0.1. Warm up: Metric and topological spaces 3
'0.2. The Banach space of continuous functions 10
'0.3. The geometry of Hilbert spaces 14
'0.4. Completeness 19
'0.5. Bounded operators 20
'0.6. Lebesgue L
p
spaces 22
'0.7. Appendix: The uniform boundedness principle 27
Part 1. Mathematical Foundations of Quantum Mechanics
Chapter 1. Hilbert spaces 31
'1.1. Hilbert spaces 31
'1.2. Orthonormal bases 33
'1.3. The projection theorem and the Riesz lemma 36
'1.4. Orthogonal sums and tensor products 38
'1.5. The C
∗
algebra of bounded linear operators 40
'1.6. Weak and strong convergence 41
'1.7. Appendix: The Stone–Weierstraß theorem 44
Chapter 2. Selfadjointness and spectrum 47
iii
iv Contents
'2.1. Some quantum mechanics 47
'2.2. Selfadjoint operators 50
'2.3. Resolvents and spectra 61
'2.4. Orthogonal sums of operators 67
'2.5. Selfadjoint extensions 68
'2.6. Appendix: Absolutely continuous functions 72
Chapter 3. The spectral theorem 75
'3.1. The spectral theorem 75
'3.2. More on Borel measures 85
'3.3. Spectral types 89
'3.4. Appendix: The Herglotz theorem 91
Chapter 4. Applications of the spectral theorem 97
'4.1. Integral formulas 97
'4.2. Commuting operators 100
'4.3. The minmax theorem 103
'4.4. Estimating eigenspaces 104
'4.5. Tensor products of operators 105
Chapter 5. Quantum dynamics 107
'5.1. The time evolution and Stone’s theorem 107
'5.2. The RAGE theorem 110
'5.3. The Trotter product formula 115
Chapter 6. Perturbation theory for selfadjoint operators 117
'6.1. Relatively bounded operators and the Kato–Rellich theorem 117
'6.2. More on compact operators 119
'6.3. Hilbert–Schmidt and trace class operators 122
'6.4. Relatively compact operators and Weyl’s theorem 128
'6.5. Strong and norm resolvent convergence 131
Part 2. Schr¨odinger Operators
Chapter 7. The free Schr¨odinger operator 139
'7.1. The Fourier transform 139
'7.2. The free Schr¨odinger operator 142
'7.3. The time evolution in the free case 144
'7.4. The resolvent and Green’s function 145
Contents v
Chapter 8. Algebraic methods 149
'8.1. Position and momentum 149
'8.2. Angular momentum 151
'8.3. The harmonic oscillator 154
Chapter 9. One dimensional Schr¨odinger operators 157
'9.1. SturmLiouville operators 157
'9.2. Weyl’s limit circle, limit point alternative 161
'9.3. Spectral transformations 168
Chapter 10. Oneparticle Schr¨odinger operators 177
'10.1. Selfadjointness and spectrum 177
'10.2. The hydrogen atom 178
'10.3. Angular momentum 181
'10.4. The eigenvalues of the hydrogen atom 184
'10.5. Nondegeneracy of the ground state 186
Chapter 11. Atomic Schr¨odinger operators 189
'11.1. Selfadjointness 189
'11.2. The HVZ theorem 191
Chapter 12. Scattering theory 197
'12.1. Abstract theory 197
'12.2. Incoming and outgoing states 200
'12.3. Schr¨odinger operators with short range potentials 202
Part 3. Appendix
Appendix A. Almost everything about Lebesgue integration 209
'A.1. Borel measures in a nut shell 209
'A.2. Extending a premasure to a measure 213
'A.3. Measurable functions 218
'A.4. The Lebesgue integral 220
'A.5. Product measures 224
'A.6. Decomposition of measures 227
'A.7. Derivatives of measures 229
Bibliography 235
Glossary of notations 237
Index 241
Preface
Overview
The present manuscript was written for my course Schr¨odinger Operators
held at the University of Vienna in Winter 1999, Summer 2002, and Summer
2005. It is supposed to give a brief but rather self contained introduction
to the mathematical methods of quantum mechanics with a view towards
applications to Schr¨odinger operators. The applications presented are highly
selective and many important and interesting items are not touched.
The ﬁrst part is a stripped down introduction to spectral theory of un
bounded operators where I try to introduce only those topics which are
needed for the applications later on. This has the advantage that you will
not get drowned in results which are never used again before you get to
the applications. In particular, I am not trying to provide an encyclope
dic reference. Nevertheless I still feel that the ﬁrst part should give you a
solid background covering all important results which are usually taken for
granted in more advanced books and research papers.
My approach is built around the spectral theorem as the central object.
Hence I try to get to it as quickly as possible. Moreover, I do not take the
detour over bounded operators but I go straight for the unbounded case. In
addition, existence of spectral measures is established via the Herglotz rather
than the Riesz representation theorem since this approach paves the way for
an investigation of spectral types via boundary values of the resolvent as the
spectral parameter approaches the real line.
vii
viii Preface
The second part starts with the free Schr¨odinger equation and computes
the free resolvent and time evolution. In addition, I discuss position, mo
mentum, and angular momentum operators via algebraic methods. This is
usually found in any physics textbook on quantum mechanics, with the only
diﬀerence that I include some technical details which are usually not found
there. Furthermore, I compute the spectrum of the hydrogen atom, again
I try to provide some mathematical details not found in physics textbooks.
Further topics are nondegeneracy of the ground state, spectra of atoms (the
HVZ theorem) and scattering theory.
Prerequisites
I assume some previous experience with Hilbert spaces and bounded
linear operators which should be covered in any basic course on functional
analysis. However, while this assumption is reasonable for mathematics
students, it might not always be for physics students. For this reason there
is a preliminary chapter reviewing all necessary results (including proofs).
In addition, there is an appendix (again with proofs) providing all necessary
results from measure theory.
Readers guide
There is some intentional overlap between Chapter 0, Chapter 1 and
Chapter 2. Hence, provided you have the necessary background, you can
start reading in Chapter 1 or even Chapter 2. Chapters 2, 3 are key chapters
and you should study them in detail (except for Section 2.5 which can be
skipped on ﬁrst reading). Chapter 4 should give you an idea of how the
spectral theorem is used. You should have a look at (e.g.) the ﬁrst section
and you can come back to the remaining ones as needed. Chapter 5 contains
two key results from quantum dynamics, Stone’s theorem and the RAGE
theorem. In particular the RAGE theorem shows the connections between
long time behavior and spectral types. Finally, Chapter 6 is again of central
importance and should be studied in detail.
The chapters in the second part are mostly independent of each others
except for the ﬁrst one, Chapter 7, which is a prerequisite for all others
except for Chapter 9.
If you are interested in one dimensional models (SturmLiouville equa
tions), Chapter 9 is all you need.
If you are interested in atoms, read Chapter 7, Chapter 10, and Chap
ter 11. In particular, you can skip the separation of variables (Sections 10.3
Preface ix
and 10.4, which require Chapter 9) method for computing the eigenvalues of
the Hydrogen atom if you are happy with the fact that there are countably
many which accumulate at the bottom of the continuous spectrum.
If you are interested in scattering theory, read Chapter 7, the ﬁrst two
sections of Chapter 10, and Chapter 12. Chapter 5 is one of the key prereq
uisites in this case.
Availability
It is available from
http://www.mat.univie.ac.at/~gerald/ftp/bookschroe/
Acknowledgments
I’d like to thank Volker Enß for making his lecture notes available to me and
Wang Lanning, Maria HoﬀmannOstenhof, Zhenyou Huang, Harald Rindler,
and Karl Unterkoﬂer for pointing out errors in previous versions.
Gerald Teschl
Vienna, Austria
February, 2005
Part 0
Preliminaries
Chapter 0
A ﬁrst look at Banach
and Hilbert spaces
I assume that the reader has some basic familiarity with measure theory and func
tional analysis. For convenience, some facts needed from Banach and L
p
spaces
are reviewed in this chapter. A crash course in measure theory can be found in
the appendix. If you feel comfortable with terms like Lebesgue L
p
spaces, Banach
space, or bounded linear operator, you can skip this entire chapter. However, you
might want to at least browse through it to refresh your memory.
0.1. Warm up: Metric and topological spaces
Before we begin I want to recall some basic facts from metric and topological
spaces. I presume that you are familiar with these topics from your calculus
course. A good reference is [8].
A metric space is a space X together with a function d : X X →R
such that
(i) d(x, y) ≥ 0
(ii) d(x, y) = 0 if and only if x = y
(iii) d(x, y) = d(y, x)
(iv) d(x, z) ≤ d(x, y) +d(y, z) (triangle inequality)
If (ii) does not hold, d is called a semimetric.
Example. Euclidean space R
n
together with d(x, y) = (
¸
n
k=1
(x
k
−y
k
)
2
)
1/2
is a metric space and so is C
n
together with d(x, y) = (
¸
n
k=1
[x
k
−y
k
[
2
)
1/2
.
3
4 0. A ﬁrst look at Banach and Hilbert spaces
The set
B
r
(x) = ¦y ∈ X[d(x, y) < r¦ (0.1)
is called an open ball around x with radius r > 0. A point x of some set
U is called an interior point of U if U contains some ball around x. If x is
an interior point of U, then U is also called a neighborhood of x. A point
x is called a limit point of U if B
r
(x) ∩ (U`¦x¦) = ∅ for every ball. Note
that a limit point must not lie in U, but U contains points arbitrarily close
to x. Moreover, x is not a limit point of U if and only if it is an interior
point of the complement of U.
Example. Consider R with the usual metric and let U = (−1, 1). Then
every point x ∈ U is an interior point of U. The points ±1 are limit points
of U.
A set consisting only of interior points is called open. The family of
open sets O satisﬁes the following properties
(i) ∅, X ∈ O
(ii) O
1
, O
2
∈ O implies O
1
∩ O
2
∈ O
(iii) ¦O
α
¦ ⊆ O implies
¸
α
O
α
∈ O
That is, O is closed under ﬁnite intersections and arbitrary unions.
In general, a space X together with a family of sets O, the open sets,
satisfying (i)–(iii) is called a topological space. The notions of interior
point, limit point, and neighborhood carry over to topological spaces if we
replace open ball by open set.
There are usually diﬀerent choices for the topology. Two usually not
very interesting examples are the trivial topology O = ¦∅, X¦ and the
discrete topology O = P(X) (the powerset of X). Given two topologies
O
1
and O
2
on X, O
1
is called weaker (or coarser) than O
2
if and only if
O
1
⊆ O
2
.
Example. Note that diﬀerent metrics can give rise to the same topology.
For example, we can equip R
n
(or C
n
) with the Euclidean distance as before,
or we could also use
˜
d(x, y) =
n
¸
k=1
[x
k
−y
k
[ (0.2)
Since
1
√
n
n
¸
k=1
[x
k
[ ≤
n
¸
k=1
[x
k
[
2
≤
n
¸
k=1
[x
k
[ (0.3)
shows B
r/
√
n
((x, y)) ⊆
˜
B
r
((x, y)) ⊆ B
r
((x, y)), where B,
˜
B are balls com
puted using d,
˜
d, respectively. Hence the topology is the same for both
metrics.
0.1. Warm up: Metric and topological spaces 5
Example. We can always replace a metric d by the bounded metric
˜
d(x, y) =
d(x, y)
1 +d(x, y)
(0.4)
without changing the topology.
Every subspace Y of a topological space X becomes a topological space
of its own if we call O ⊆ Y open if there is some open set
˜
O ⊆ X such that
O =
˜
O ∩ Y (induced topology).
Example. The set (0, 1] ⊆ R is not open in the topology of X = R, but it is
open in the incuded topology when considered as a subset of Y = [−1, 1].
A family of open sets B ⊆ O is called a base for the topology if for each
x and each neighborhood U(x), there is some set O ∈ B with x ∈ O ⊆ U.
Since O =
¸
x∈O
U(x) we have
Lemma 0.1. If B ⊆ O is a base for the topology, then every open set can
be written as a union of elements from B.
If there exists a countable base, then X is called second countable.
Example. By construction the open balls B
1/n
(x) are a base for the topol
ogy in a metric space. In the case of R
n
(or C
n
) it even suﬃces to take balls
with rational center and hence R
n
(and C
n
) are second countable.
A topological space is called Hausdorﬀ space if for two diﬀerent points
there are always two disjoint neighborhoods.
Example. Any metric space is a Hausdorﬀ space: Given two diﬀerent
points x and y the balls B
d/2
(x) and B
d/2
(y), where d = d(x, y) > 0, are
disjoint neighborhoods (a semimetric space will not be Hausdorﬀ).
The complement of an open set is called a closed set. It follows from
de Morgan’s rules that the family of closed sets ( satisﬁes
(i) ∅, X ∈ (
(ii) C
1
, C
2
∈ ( implies C
1
∪ C
2
∈ (
(iii) ¦C
α
¦ ⊆ ( implies
¸
α
C
α
∈ (
That is, closed sets are closed under ﬁnite unions and arbitrary intersections.
The smallest closed set containing a given set U is called the closure
U =
¸
C∈C,U⊆C
C, (0.5)
and the largest open set contained in a given set U is called the interior
U
◦
=
¸
O∈O,O⊆U
O. (0.6)
6 0. A ﬁrst look at Banach and Hilbert spaces
It is straightforward to check that
Lemma 0.2. Let X be a topological space, then the interior of U is the set
of all interior points of U and the closure of U is the set of all limit points
of U.
A sequence (x
n
)
∞
n=1
⊆ X is said to converge to some point x ∈ X if
d(x, x
n
) → 0. We write lim
n→∞
x
n
= x as usual in this case. Clearly the
limit is unique if it exists (this is not true for a semimetric).
Every convergent sequence is a Cauchy sequence, that is, for every
ε > 0 there is some N ∈ N such that
d(x
n
, x
m
) ≤ ε n, m ≥ N. (0.7)
If the converse is also true, that is, if every Cauchy sequence has a limit,
then X is called complete.
Example. Both R
n
and C
n
are complete metric spaces.
A point x is clearly a limit point of U if and only if there is some sequence
x
n
∈ U converging to x. Hence
Lemma 0.3. A closed subset of a complete metric space is again a complete
metric space.
Note that convergence can also be equivalently formulated in terms of
topological terms: A sequence x
n
converges to x if and only if for every
neighborhood U of x there is some N ∈ N such that x
n
∈ U for n ≥ N. In
a Hausdorﬀ space the limit is unique.
A metric space is called separable if it contains a countable dense set.
A set U is called dense, if its closure is all of X, that is if U = X.
Lemma 0.4. Let X be a separable metric space. Every subset of X is again
separable.
Proof. Let A = ¦x
n
¦
n∈N
be a dense set in X. The only problem is that
A∩Y might contain no elements at all. However, some elements of A must
be at least arbitrarily close: Let J ⊆ N
2
be the set of all pairs (n, m) for
which B
1/m
(x
n
) ∩ Y = ∅ and choose some y
n,m
∈ B
1/m
(x
n
) ∩ Y for all
(n, m) ∈ J. Then B = ¦y
n,m
¦
(n,m)∈J
⊆ Y is countable. To see that B is
dense choose y ∈ Y . Then there is some sequence x
n
k
with d(x
n
k
, y) < 1/4.
Hence (n
k
, k) ∈ J and d(y
n
k
,k
, y) ≤ d(y
n
k
,k
, x
n
k
) +d(x
n
k
, y) ≤ 2/k → 0.
A function between metric spaces X and Y is called continuous at a
point x ∈ X if for every ε > 0 we can ﬁnd a δ > 0 such that
d
Y
(f(x), f(y)) ≤ ε if d
X
(x, y) < δ. (0.8)
If f is continuous at every point it is called continuous.
0.1. Warm up: Metric and topological spaces 7
Lemma 0.5. Let X be a metric space. The following are equivalent
(i) f is continuous at x (i.e, (0.8) holds).
(ii) f(x
n
) → f(x) whenever x
n
→ x
(iii) For every neighborhood V of f(x), f
−1
(V ) is a neighborhood of x.
Proof. (i) ⇒ (ii) is obvious. (ii) ⇒ (iii): If (iii) does not hold there is
a neighborhood V of f(x) such that B
δ
(x) ⊆ f
−1
(V ) for every δ. Hence
we can choose a sequence x
n
∈ B
1/n
(x) such that f(x
n
) ∈ f
−1
(V ). Thus
x
n
→ x but f(x
n
) → f(x). (iii) ⇒ (i): Choose V = B
ε
(f(x)) and observe
that by (iii) B
δ
(x) ⊆ f
−1
(V ) for some δ.
The last item implies that f is continuous if and only if the inverse image
of every open (closed) set is again open (closed).
Note: In a topological space, (iii) is used as deﬁnition for continuity.
However, in general (ii) and (iii) will no longer be equivalent unless one uses
generalized sequences, so called nets, where the index set N is replaced by
arbitrary directed sets.
If X and X are metric spaces then X Y together with
d((x
1
, y
1
), (x
2
, y
2
)) = d
X
(x
1
, x
2
) +d
Y
(y
1
, y
2
) (0.9)
is a metric space. A sequence (x
n
, y
n
) converges to (x, y) if and only if
x
n
→ x and y
n
→ y. In particular, the projections onto the ﬁrst (x, y) → x
respectively onto the second (x, y) → y coordinate are continuous.
In particular, by
[d(x
n
, y
n
) −d(x, y)[ ≤ d(x
n
, x) +d(y
n
, y) (0.10)
we see that d : X X →R is continuous.
Example. If we consider RR we do not get the Euclidean distance of R
2
unless we modify (0.9) as follows:
˜
d((x
1
, y
1
), (x
2
, y
2
)) =
d
X
(x
1
, x
2
)
2
+d
Y
(y
1
, y
2
)
2
. (0.11)
As noted in our previous example, the topology (and thus also conver
gence/continuity) is independent of this choice.
If X and Y are just topological spaces, the product topology is deﬁned
by calling O ⊆ X Y open if for every point (x, y) ∈ O there are open
neighborhoods U of x and V of y such that U V ⊆ O. In the case of
metric spaces this clearly agrees with the topology deﬁned via the product
metric (0.9).
A cover of a set Y ⊆ X is a family of sets ¦U
α
¦ such that Y ⊆
¸
α
U
α
. A
cover is call open if all U
α
are open. A subset of ¦U
α
¦ is called a subcover.
8 0. A ﬁrst look at Banach and Hilbert spaces
A subset K ⊂ X is called compact if every open cover has a ﬁnite
subcover.
Lemma 0.6. A topological space is compact if and only if it has the ﬁnite
intersection property: The intersection of a family of closed sets is empty
if and only if the intersection of some ﬁnite subfamily is empty.
Proof. By taking complements, to every family of open sets there is a cor
responding family of closed sets and vice versa. Moreover, the open sets
are a cover if and only if the corresponding closed sets have empty intersec
tion.
A subset K ⊂ X is called sequentially compact if every sequence has
a convergent subsequence.
Lemma 0.7. Let X be a topological space.
(i) The continuous image of a compact set is compact.
(ii) Every closed subset of a compact set is compact.
(iii) If X is Hausdorﬀ, any compact set is closed.
(iv) The product of compact sets is compact.
(v) A compact set is also sequentially compact.
Proof. (i) Just observe that if ¦O
α
¦ is an open cover for f(Y ), then ¦f
−1
(O
α
)¦
is one for Y .
(ii) Let ¦O
α
¦ be an open cover for the closed subset Y . Then ¦O
α
¦ ∪
¦X`Y ¦ is an open cover for X.
(iii) Let Y ⊆ X be compact. We show that X`Y is open. Fix x ∈ X`Y
(if Y = X there is nothing to do). By the deﬁnition of Hausdorﬀ, for
every y ∈ Y there are disjoint neighborhoods V (y) of y and U
y
(x) of x. By
compactness of Y , there are y
1
, . . . y
n
such that V (y
j
) cover Y . But then
U(x) =
¸
n
j=1
U
y
j
(x) is a neighborhood of x which does not intersect Y .
(iv) Let ¦O
α
¦ be an open cover for X Y . For every (x, y) ∈ X Y
there is some α(x, y) such that (x, y) ∈ O
α(x,y)
. By deﬁnition of the product
topology there is some open rectangle U(x, y) V (x, y) ⊆ O
α(x,y)
. Hence
for ﬁxed x, ¦V (x, y)¦
y∈Y
is an open cover of Y . Hence there are ﬁnitely
many points y
k
(x) such V (x, y
k
(x)) cover Y . Set U(x) =
¸
k
U(x, y
k
(x)).
Since ﬁnite intersections of open sets are open, ¦U(x)¦
x∈X
is an open cover
and there are ﬁnitely many points x
j
such U(x
j
) cover X. By construction,
U(x
j
) V (x
j
, y
k
(x
j
)) ⊆ O
α(x
j
,y
k
(x
j
))
cover X Y .
(v) Let x
n
be a sequence which has no convergent subsequence. Then
K = ¦x
n
¦ has no limit points and is hence compact by (ii). For every n
0.1. Warm up: Metric and topological spaces 9
there is a ball B
εn
(x
n
) which contains only ﬁnitely many elements of K.
However, ﬁnitely many suﬃce to cover K, a contradiction.
In a metric space compact and sequentially compact are equivalent.
Lemma 0.8. Let X be a metric space. Then a subset is compact if and only
if it is sequentially compact.
Proof. First of all note that every cover of open balls with ﬁxed radius
ε > 0 has a ﬁnite subcover. Since if this were false we could construct a
sequence x
n
∈ X`
¸
n−1
m=1
B
ε
(x
m
) such that d(x
n
, x
m
) > ε for m < n.
In particular, we are done if we can show that for every open cover
¦O
α
¦ there is some ε > 0 such that for every x we have B
ε
(x) ⊆ O
α
for
some α = α(x). Indeed, choosing ¦x
k
¦
n
k=1
such that B
ε
(x
k
) is a cover, we
have that O
α(x
k
)
is a cover as well.
So it remains to show that there is such an ε. If there were none, for
every ε > 0 there must be an x such that B
ε
(x) ⊆ O
α
for every α. Choose
ε =
1
n
and pick a corresponding x
n
. Since X is sequentially compact, it is no
restriction to assume x
n
converges (after maybe passing to a subsequence).
Let x = limx
n
, then x lies in some O
α
and hence B
ε
(x) ⊆ O
α
. But choosing
n so large that
1
n
<
ε
2
and d(x
n
, x) <
ε
2
we have B
1/n
(x
n
) ⊆ B
ε
(x) ⊆ O
α
contradicting our assumption.
Please also recall the HeineBorel theorem:
Theorem 0.9 (HeineBorel). In R
n
(or C
n
) a set is compact if and only if
it is bounded and closed.
Proof. By Lemma 0.7 (ii) and (iii) it suﬃces to show that a closed interval
in I ⊆ R is compact. Moreover, by Lemma 0.8 it suﬃces to show that
every sequence in I = [a, b] has a convergent subsequence. Let x
n
be our
sequence and divide I = [a,
a+b
2
] ∪ [
a+b
2
]. Then at least one of these two
intervals, call it I
1
, contains inﬁnitely many elements of our sequence. Let
y
1
= x
n
1
be the ﬁrst one. Subdivide I
1
and pick y
2
= x
n
2
, with n
2
> n
1
as
before. Proceeding like this we obtain a Cauchy sequence y
n
(note that by
construction I
n+1
⊆ I
n
and hence [y
n
−y
m
[ ≤
b−a
n
for m ≥ n).
A topological space is called locally compact if every point has a com
pact neighborhood.
Example. R
n
is locally compact.
The distance between a point x ∈ X and a subset Y ⊆ X is
dist(x, Y ) = inf
y∈Y
d(x, y). (0.12)
10 0. A ﬁrst look at Banach and Hilbert spaces
Note that x ∈ Y if and only if dist(x, Y ) = 0.
Lemma 0.10. Let X be a metric space, then
[dist(x, Y ) −dist(z, Y )[ ≤ dist(x, z). (0.13)
In particular, x → dist(x, Y ) is continuous.
Proof. Taking the inﬁmum in the triangle inequality d(x, y) ≤ d(x, z) +
d(z, y) shows dist(x, Y ) ≤ d(x, z)+dist(z, Y ). Hence dist(x, Y )−dist(z, Y ) ≤
dist(x, z). Interchanging x and z shows dist(z, Y ) − dist(x, Y ) ≤ dist(x, z).
Lemma 0.11 (Urysohn). Suppose C
1
and C
2
are disjoint closed subsets of
a metric space X. Then there is a continuous function f : X → [0, 1] such
that f is zero on C
1
and one on C
2
.
If X is locally compact and U is compact, one can choose f with compact
support.
Proof. To prove the ﬁrst claim set f(x) =
dist(x,C
2
)
dist(x,C
1
)+dist(x,C
2
)
. For the
second claim, observe that there is an open set O such that O is compact
and C
1
⊂ O ⊂ O ⊂ X`C
2
. In fact, for every x, there is a ball B
ε
(x) such
that B
ε
(x) is compact and B
ε
(x) ⊂ X`C
2
. Since U is compact, ﬁnitely
many of them cover C
1
and we can choose the union of those balls to be O.
Now replace C
2
by X`O.
Note that Urysohn’s lemma implies that a metric space is normal, that
is, for any two disjoint closed sets C
1
and C
2
, there are disjoint open sets
O
1
and O
2
such that C
j
⊆ O
j
, j = 1, 2. In fact, choose f as in Urysohn’s
lemma and set O
1
= f
−1
([0, 1/2)) respectively O
2
= f
−1
((1/2, 1]).
0.2. The Banach space of continuous functions
Now let us have a ﬁrst look at Banach spaces by investigating set of contin
uous functions C(I) on a compact interval I = [a, b] ⊂ R. Since we want to
handle complex models, we will always consider complex valued functions!
One way of declaring a distance, wellknown from calculus, is the max
imum norm:
f(x) −g(x)
∞
= max
x∈I
[f(x) −g(x)[. (0.14)
It is not hard to see that with this deﬁnition C(I) becomes a normed linear
space:
A normed linear space X is a vector space X over C (or R) with a
realvalued function (the norm) . such that
• f ≥ 0 for all f ∈ X and f = 0 if and only if f = 0,
0.2. The Banach space of continuous functions 11
• λf = [λ[ f for all λ ∈ C and f ∈ X, and
• f +g ≤ f +g for all f, g ∈ X (triangle inequality).
From the triangle inequality we also get the inverse triangle inequality
(Problem 0.1)
[f −g[ ≤ f −g. (0.15)
Once we have a norm, we have a distance d(f, g) = f−g and hence we
know when a sequence of vectors f
n
converges to a vector f. We will write
f
n
→ f or lim
n→∞
f
n
= f, as usual, in this case. Moreover, a mapping
F : X → Y between to normed spaces is called continuous if f
n
→ f
implies F(f
n
) → F(f). In fact, it is not hard to see that the norm, vector
addition, and multiplication by scalars are continuous (Problem 0.2).
In addition to the concept of convergence we have also the concept of
a Cauchy sequence and hence the concept of completeness: A normed
space is called complete if every Cauchy sequence has a limit. A complete
normed space is called a Banach space.
Example. The space
1
(N) of all sequences a = (a
j
)
∞
j=1
for which the norm
a
1
=
∞
¸
j=1
[a
j
[ (0.16)
is ﬁnite, is a Banach space.
To show this, we need to verify three things: (i)
1
(N) is a Vector space,
that is closed under addition and scalar multiplication (ii) .
1
satisﬁes the
three requirements for a norm and (iii)
1
(N) is complete.
First of all observe
k
¸
j=1
[a
j
+b
j
[ ≤
k
¸
j=1
[a
j
[ +
k
¸
j=1
[b
j
[ ≤ a
1
+b
1
(0.17)
for any ﬁnite k. Letting k → ∞ we conclude that
1
(N) is closed under
addition and that the triangle inequality holds. That
1
(N) is closed under
scalar multiplication and the two other properties of a norm are straight
forward. It remains to show that
1
(N) is complete. Let a
n
= (a
n
j
)
∞
j=1
be
a Cauchy sequence, that is, for given ε > 0 we can ﬁnd an N
ε
such that
a
m
−a
n

1
≤ ε for m, n ≥ N
ε
. This implies in particular [a
m
j
−a
n
j
[ ≤ ε for
any ﬁxed j. Thus a
n
j
is a Cauchy sequence for ﬁxed j and by completeness
of C has a limit: lim
n→∞
a
n
j
= a
j
. Now consider
k
¸
j=1
[a
m
j
−a
n
j
[ ≤ ε (0.18)
12 0. A ﬁrst look at Banach and Hilbert spaces
and take m → ∞:
k
¸
j=1
[a
j
−a
n
j
[ ≤ ε. (0.19)
Since this holds for any ﬁnite k we even have a−a
n

1
≤ ε. Hence (a−a
n
) ∈
1
(N) and since a
n
∈
1
(N) we ﬁnally conclude a = a
n
+(a−a
n
) ∈
1
(N).
Example. The space
∞
(N) of all bounded sequences a = (a
j
)
∞
j=1
together
with the norm
a
∞
= sup
j∈N
[a
j
[ (0.20)
is a Banach space (Problem 0.3).
Now what about convergence in this space? A sequence of functions
f
n
(x) converges to f if and only if
lim
n→∞
f −f
n
 = lim
n→∞
sup
x∈I
[f
n
(x) −f(x)[ = 0. (0.21)
That is, in the language of real analysis, f
n
converges uniformly to f. Now
let us look at the case where f
n
is only a Cauchy sequence. Then f
n
(x) is
clearly a Cauchy sequence of real numbers for any ﬁxed x ∈ I. In particular,
by completeness of C, there is a limit f(x) for each x. Thus we get a limiting
function f(x). Moreover, letting m → ∞ in
[f
m
(x) −f
n
(x)[ ≤ ε ∀m, n > N
ε
, x ∈ I (0.22)
we see
[f(x) −f
n
(x)[ ≤ ε ∀n > N
ε
, x ∈ I, (0.23)
that is, f
n
(x) converges uniformly to f(x). However, up to this point we
don’t know whether it is in our vector space C(I) or not, that is, whether
it is continuous or not. Fortunately, there is a wellknown result from real
analysis which tells us that the uniform limit of continuous functions is again
continuous. Hence f(x) ∈ C(I) and thus every Cauchy sequence in C(I)
converges. Or, in other words
Theorem 0.12. C(I) with the maximum norm is a Banach space.
Next we want to know if there is a basis for C(I). In order to have only
countable sums, we would even prefer a countable basis. If such a basis
exists, that is, if there is a set ¦u
n
¦ ⊂ X of linearly independent vectors
such that every element f ∈ X can be written as
f =
¸
n
c
n
u
n
, c
n
∈ C, (0.24)
then the span span¦u
n
¦ (the set of all ﬁnite linear combinations) of ¦u
n
¦ is
dense in X. A set whose span is dense is called total and if we have a total
set, we also have a countable dense set (consider only linear combinations
0.2. The Banach space of continuous functions 13
with rational coeﬃcients – show this). A normed linear space containing a
countable dense set is called separable.
Example. The Banach space
1
(N) is separable. In fact, the set of vectors
δ
n
, with δ
n
n
= 1 and δ
n
m
= 0, n = m is total: Let a ∈
1
(N) be given and set
a
n
=
¸
n
k=1
a
k
δ
k
, then
a −a
n

1
=
∞
¸
j=n+1
[a
j
[ → 0 (0.25)
since a
n
j
= a
j
for 1 ≤ j ≤ n and a
n
j
= 0 for j > n.
Luckily this is also the case for C(I):
Theorem 0.13 (Weierstraß). Let I be a compact interval. Then the set of
polynomials is dense in C(I).
Proof. Let f(x) ∈ C(I) be given. By considering f(x) − f(a) + (f(b) −
f(a))(x −b) it is no loss to assume that f vanishes at the boundary points.
Moreover, without restriction we only consider I = [
−1
2
,
1
2
] (why?).
Now the claim follows from the lemma below using
u
n
(x) =
1
I
n
(1 −x
2
)
n
, (0.26)
where
I
n
=
1
−1
(1 −x
2
)
n
dx =
n!
1
2
(
1
2
+ 1) (
1
2
+n)
=
√
π
Γ(1 +n)
Γ(
3
2
+n)
=
π
n
(1 +O(
1
n
)). (0.27)
(Remark: The integral is known as Beta function and the asymptotics follow
from Stirling’s formula.)
Lemma 0.14 (Smoothing). Let u
n
(x) be a sequence of nonnegative contin
uous functions on [−1, 1] such that
x≤1
u
n
(x)dx = 1 and
δ≤x≤1
u
n
(x)dx → 0, δ > 0. (0.28)
(In other words, u
n
has mass one and concentrates near x = 0 as n → ∞.)
Then for every f ∈ C[−
1
2
,
1
2
] which vanishes at the endpoints, f(−
1
2
) =
f(
1
2
) = 0, we have that
f
n
(x) =
1/2
−1/2
u
n
(x −y)f(y)dy (0.29)
converges uniformly to f(x).
14 0. A ﬁrst look at Banach and Hilbert spaces
Proof. Since f is uniformly continuous, for given ε we can ﬁnd a δ (inde
pendent of x) such that [f(x)−f(y)[ ≤ ε whenever [x−y[ ≤ δ. Moreover, we
can choose n such that
δ≤y≤1
u
n
(y)dy ≤ ε. Now abbreviate M = max f
and note
[f(x)−
1/2
−1/2
u
n
(x−y)f(x)dy[ = [f(x)[ [1−
1/2
−1/2
u
n
(x−y)dy[ ≤ Mε. (0.30)
In fact, either the distance of x to one of the boundary points ±
1
2
is smaller
than δ and hence [f(x)[ ≤ ε or otherwise the diﬀerence between one and the
integral is smaller than ε.
Using this we have
[f
n
(x) −f(x)[ ≤
1/2
−1/2
u
n
(x −y)[f(y) −f(x)[dy +Mε
≤
y≤1/2,x−y≤δ
u
n
(x −y)[f(y) −f(x)[dy
+
y≤1/2,x−y≥δ
u
n
(x −y)[f(y) −f(x)[dy +Mε
= ε + 2Mε +Mε = (1 + 3M)ε, (0.31)
which proves the claim.
Note that f
n
will be as smooth as u
n
, hence the title smoothing lemma.
The same idea is used to approximate noncontinuous functions by smooth
ones (of course the convergence will no longer be uniform in this case).
Corollary 0.15. C(I) is separable.
The same is true for
1
(N), but not for
∞
(N) (Problem 0.4)!
Problem 0.1. Show that [f −g[ ≤ f −g.
Problem 0.2. Show that the norm, vector addition, and multiplication by
scalars are continuous. That is, if f
n
→ f, g
n
→ g, and λ
n
→ λ then
f
n
 → f, f
n
+g
n
→ f +g, and λ
n
g
n
→ λg.
Problem 0.3. Show that
∞
(N) is a Banach space.
Problem 0.4. Show that
∞
(N) is not separable (Hint: Consider sequences
which take only the value one and zero. How many are there? What is the
distance between two such sequences?).
0.3. The geometry of Hilbert spaces
So it looks like C(I) has all the properties we want. However, there is
still one thing missing: How should we deﬁne orthogonality in C(I)? In
0.3. The geometry of Hilbert spaces 15
Euclidean space, two vectors are called orthogonal if their scalar product
vanishes, so we would need a scalar product:
Suppose H is a vector space. A map '., ..` : H H → C is called skew
linear form if it is conjugate linear in the ﬁrst and linear in the second
argument, that is,
'α
1
f
1
+α
2
f
2
, g` = α
∗
1
'f
1
, g` +α
∗
2
'f
2
, g`
'f, α
1
g
1
+α
2
g
2
` = α
1
'f, g
1
` +α
2
'f, g
2
`
, α
1
, α
2
∈ C, (0.32)
where ‘∗’ denotes complex conjugation. A skew linear form satisfying the
requirements
(i) 'f, f` > 0 for f = 0 (positive deﬁnite)
(ii) 'f, g` = 'g, f`
∗
(symmetry)
is called inner product or scalar product. Associated with every scalar
product is a norm
f =
'f, f`. (0.33)
The pair (H, '., ..`) is called inner product space. If H is complete it is
called a Hilbert space.
Example. Clearly C
n
with the usual scalar product
'a, b` =
n
¸
j=1
a
∗
j
b
j
(0.34)
is a (ﬁnite dimensional) Hilbert space.
Example. A somewhat more interesting example is the Hilbert space
2
(N),
that is, the set of all sequences
(a
j
)
∞
j=1
∞
¸
j=1
[a
j
[
2
< ∞
¸
(0.35)
with scalar product
'a, b` =
∞
¸
j=1
a
∗
j
b
j
. (0.36)
(Show that this is in fact a separable Hilbert space! Problem 0.5)
Of course I still owe you a proof for the claim that
'f, f` is indeed a
norm. Only the triangle inequality is nontrivial which will follow from the
CauchySchwarz inequality below.
A vector f ∈ H is called normalized or unit vector if f = 1. Two
vectors f, g ∈ H are called orthogonal or perpendicular (f ⊥ g) if 'f, g` =
0 and parallel if one is a multiple of the other.
16 0. A ﬁrst look at Banach and Hilbert spaces
For two orthogonal vectors we have the Pythagorean theorem:
f +g
2
= f
2
+g
2
, f ⊥ g, (0.37)
which is one line of computation.
Suppose u is a unit vector, then the projection of f in the direction of
u is given by
f
= 'u, f`u (0.38)
and f
⊥
deﬁned via
f
⊥
= f −'u, f`u (0.39)
is perpendicular to u since 'u, f
⊥
` = 'u, f −'u, f`u` = 'u, f` −'u, f`'u, u` =
0.
f
f
f
⊥
u
I
I
f
f
f
f
fw
Taking any other vector parallel to u it is easy to see
f −αu
2
= f
⊥
+ (f
−αu)
2
= f
⊥

2
+['u, f` −α[
2
(0.40)
and hence f
= 'u, f`u is the unique vector parallel to u which is closest to
f.
As a ﬁrst consequence we obtain the CauchySchwarzBunjakowski
inequality:
Theorem 0.16 (CauchySchwarzBunjakowski). Let H
0
be an inner product
space, then for every f, g ∈ H
0
we have
['f, g`[ ≤ f g (0.41)
with equality if and only if f and g are parallel.
Proof. It suﬃces to prove the case g = 1. But then the claim follows
from f
2
= ['g, f`[
2
+f
⊥

2
.
Note that the CauchySchwarz inequality entails that the scalar product
is continuous in both variables, that is, if f
n
→ f and g
n
→ g we have
'f
n
, g
n
` → 'f, g`.
As another consequence we infer that the map . is indeed a norm.
f +g
2
= f
2
+'f, g` +'g, f` +g
2
≤ (f +g)
2
. (0.42)
But let us return to C(I). Can we ﬁnd a scalar product which has the
maximum norm as associated norm? Unfortunately the answer is no! The
0.3. The geometry of Hilbert spaces 17
reason is that the maximum norm does not satisfy the parallelogram law
(Problem 0.7).
Theorem 0.17 (Jordanvon Neumann). A norm is associated with a scalar
product if and only if the parallelogram law
f +g
2
+f −g
2
= 2f
2
+ 2g
2
(0.43)
holds.
In this case the scalar product can be recovered from its norm by virtue
of the polarization identity
'f, g` =
1
4
f +g
2
−f −g
2
+ if −ig
2
−if + ig
2
. (0.44)
Proof. If an inner product space is given, veriﬁcation of the parallelogram
law and the polarization identity is straight forward (Problem 0.6).
To show the converse, we deﬁne
s(f, g) =
1
4
f +g
2
−f −g
2
+ if −ig
2
−if + ig
2
. (0.45)
Then s(f, f) = f
2
and s(f, g) = s(g, f)
∗
are straightforward to check.
Moreover, another straightforward computation using the parallelogram law
shows
s(f, g) +s(f, h) = 2s(f,
g +h
2
). (0.46)
Now choosing h = 0 (and using s(f, 0) = 0) shows s(f, g) = 2s(f,
g
2
) and
thus s(f, g) + s(f, h) = s(f, g + h). Furthermore, by induction we infer
m
2
n
s(f, g) = s(f,
m
2
n
g), that is λs(f, g) = s(f, λg) for every positive rational λ.
By continuity (check this!) this holds for all λ > 0 and s(f, −g) = −s(f, g)
respectively s(f, ig) = i s(f, g) ﬁnishes the proof.
Note that the parallelogram law and the polarization identity even hold
for skew linear forms (Problem 0.6).
But how do we deﬁne a scalar product on C(I)? One possibility is
'f, g` =
b
a
f
∗
(x)g(x)dx. (0.47)
The corresponding inner product space is denoted by L
2
cont
(I). Note that
we have
f ≤
[b −a[f
∞
(0.48)
and hence the maximum norm is stronger than the L
2
cont
norm.
Suppose we have two norms .
1
and .
2
on a space X. Then .
2
is
said to be stronger than .
1
if there is a constant m > 0 such that
f
1
≤ mf
2
. (0.49)
18 0. A ﬁrst look at Banach and Hilbert spaces
It is straightforward to check that
Lemma 0.18. If .
2
is stronger than .
1
, then any .
2
Cauchy sequence
is also a .
1
Cauchy sequence.
Hence if a function F : X → Y is continuous in (X, .
1
) it is also
continuos in (X, .
2
) and if a set is dense in (X, .
2
) it is also dense in
(X, .
1
).
In particular, L
2
cont
is separable. But is it also complete? Unfortunately
the answer is no:
Example. Take I = [0, 2] and deﬁne
f
n
(x) =
0, 0 ≤ x ≤ 1 −
1
n
1 +n(x −1), 1 −
1
n
≤ x ≤ 1
1, 1 ≤ x ≤ 2
(0.50)
then f
n
(x) is a Cauchy sequence in L
2
cont
, but there is no limit in L
2
cont
!
Clearly the limit should be the step function which is 0 for 0 ≤ x < 1 and 1
for 1 ≤ x ≤ 2, but this step function is discontinuous (Problem 0.8)!
This shows that in inﬁnite dimensional spaces diﬀerent norms will give
raise to diﬀerent convergent sequences! In fact, the key to solving prob
lems in inﬁnite dimensional spaces is often ﬁnding the right norm! This is
something which cannot happen in the ﬁnite dimensional case.
Theorem 0.19. If X is a ﬁnite dimensional case, then all norms are equiv
alent. That is, for given two norms .
1
and .
2
there are constants m
1
and m
2
such that
1
m
2
f
1
≤ f
2
≤ m
1
f
1
. (0.51)
Proof. Clearly we can choose a basis u
j
, 1 ≤ j ≤ n, and assume that .
2
is
the usual Euclidean norm, 
¸
j
α
j
u
j

2
2
=
¸
j
[α
j
[
2
. Let f =
¸
j
α
j
u
j
, then
by the triangle and Cauchy Schwartz inequalities
f
1
≤
¸
j
[α
j
[u
j

1
≤
¸
j
u
j

2
1
f
2
(0.52)
and we can choose m
2
=
¸
j
u
j

1
.
In particular, if f
n
is convergent with respect to .
2
it is also convergent
with respect to .
1
. Thus .
1
is continuous with respect to .
2
and attains
its minimum m > 0 on the unit sphere (which is compact by the HeineBorel
theorem). Now choose m
1
= 1/m.
Problem 0.5. Show that
2
(N) is a separable Hilbert space.
0.4. Completeness 19
Problem 0.6. Let s(f, g) be a skew linear form and p(f) = s(f, f) the
associated quadratic form. Prove the parallelogram law
p(f +g) +p(f −g) = 2p(f) + 2p(g) (0.53)
and the polarization identity
s(f, g) =
1
4
(p(f +g) −p(f −g) + i p(f −ig) −i p(f + ig)) . (0.54)
Problem 0.7. Show that the maximum norm (on C[0, 1]) does not satisfy
the parallelogram law.
Problem 0.8. Prove the claims made about f
n
, deﬁned in (0.50), in the
last example.
0.4. Completeness
Since L
2
cont
is not complete, how can we obtain a Hilbert space out of it?
Well the answer is simple: take the completion.
If X is a (incomplete) normed space, consider the set of all Cauchy
sequences
˜
X. Call two Cauchy sequences equivalent if their diﬀerence con
verges to zero and denote by
¯
X the set of all equivalence classes. It is easy
to see that
¯
X (and
˜
X) inherit the vector space structure from X. Moreover,
Lemma 0.20. If x
n
is a Cauchy sequence, then x
n
 converges.
Consequently the norm of a Cauchy sequence (x
n
)
∞
n=1
can be deﬁned by
(x
n
)
∞
n=1
 = lim
n→∞
x
n
 and is independent of the equivalence class (show
this!). Thus
¯
X is a normed space (
˜
X is not! why?).
Theorem 0.21.
¯
X is a Banach space containing X as a dense subspace if
we identify x ∈ X with the equivalence class of all sequences converging to
x.
Proof. (Outline) It remains to show that
¯
X is complete. Let ξ
n
= [(x
n,j
)
∞
j=1
]
be a Cauchy sequence in
¯
X. Then it is not hard to see that ξ = [(x
j,j
)
∞
j=1
]
is its limit.
Let me remark that the completion X is unique. More precisely any
other complete space which contains X as a dense subset is isomorphic to
X. This can for example be seen by showing that the identity map on X
has a unique extension to X (compare Theorem 0.24 below).
In particular it is no restriction to assume that a normed linear space
or an inner product space is complete. However, in the important case
of L
2
cont
it is somewhat inconvenient to work with equivalence classes of
Cauchy sequences and hence we will give a diﬀerent characterization using
the Lebesgue integral later.
20 0. A ﬁrst look at Banach and Hilbert spaces
0.5. Bounded operators
A linear map A between two normed spaces X and Y will be called a (lin
ear) operator
A : D(A) ⊆ X → Y. (0.55)
The linear subspace D(A) on which A is deﬁned, is called the domain of A
and is usually required to be dense. The kernel
Ker(A) = ¦f ∈ D(A)[Af = 0¦ (0.56)
and range
Ran(A) = ¦Af[f ∈ D(A)¦ = AD(A) (0.57)
are deﬁned as usual. The operator A is called bounded if the following
operator norm
A = sup
f
X
=1
Af
Y
(0.58)
is ﬁnite.
The set of all bounded linear operators from X to Y is denoted by
L(X, Y ). If X = Y we write L(X, X) = L(X).
Theorem 0.22. The space L(X, Y ) together with the operator norm (0.58)
is a normed space. It is a Banach space if Y is.
Proof. That (0.58) is indeed a norm is straightforward. If Y is complete and
A
n
is a Cauchy sequence of operators, then A
n
f converges to an element
g for every f. Deﬁne a new operator A via Af = g. By continuity of
the vector operations, A is linear and by continuity of the norm Af =
lim
n→∞
A
n
f ≤ (lim
n→∞
A
n
)f it is bounded. Furthermore, given
ε > 0 there is some N such that A
n
− A
m
 ≤ ε for n, m ≥ N and thus
A
n
f −A
m
f ≤ εf. Taking the limit m → ∞ we see A
n
f −Af ≤ εf,
that is A
n
→ A.
By construction, a bounded operator is Lipschitz continuous
Af
Y
≤ Af
X
(0.59)
and hence continuous. The converse is also true
Theorem 0.23. An operator A is bounded if and only if it is continuous.
Proof. Suppose A is continuous but not bounded. Then there is a sequence
of unit vectors u
n
such that Au
n
 ≥ n. Then f
n
=
1
n
u
n
converges to 0 but
Af
n
 ≥ 1 does not converge to 0.
Moreover, if A is bounded and densely deﬁned, it is no restriction to
assume that it is deﬁned on all of X.
0.5. Bounded operators 21
Theorem 0.24. Let A ∈ L(X, Y ) and let Y be a Banach space. If D(A)
is dense, there is a unique (continuous) extension of A to X, which has the
same norm.
Proof. Since a bounded operator maps Cauchy sequences to Cauchy se
quences, this extension can only be given by
Af = lim
n→∞
Af
n
, f
n
∈ D(A), f ∈ X. (0.60)
To show that this deﬁnition is independent of the sequence f
n
→ f, let
g
n
→ f be a second sequence and observe
Af
n
−Ag
n
 = A(f
n
−g
n
) ≤ Af
n
−g
n
 → 0. (0.61)
From continuity of vector addition and scalar multiplication it follows that
our extension is linear. Finally, from continuity of the norm we conclude
that the norm does not increase.
An operator in L(X, C) is called a bounded linear functional and the
space X
∗
= L(X, C) is called the dual space of X. A sequence f
n
is said to
converge weakly f
n
f if (f
n
) → (f) for every ∈ X
∗
.
The Banach space of bounded linear operators L(X) even has a multi
plication given by composition. Clearly this multiplication satisﬁes
(A+B)C = AC +BC, A(B +C) = AB +BC, A, B, C ∈ L(X) (0.62)
and
(AB)C = A(BC), α(AB) = (αA)B = A(αB), α ∈ C. (0.63)
Moreover, it is easy to see that we have
AB ≤ AB. (0.64)
However, note that our multiplication is not commutative (unless X is one
dimensional). We even have an identity, the identity operator I satisfying
I = 1.
A Banach space together with a multiplication satisfying the above re
quirements is called a Banach algebra. In particular, note that (0.64)
ensures that multiplication is continuous.
Problem 0.9. Show that the integral operator
(Kf)(x) =
1
0
K(x, y)f(y)dy, (0.65)
where K(x, y) ∈ C([0, 1] [0, 1]), deﬁned on D(K) = C[0, 1] is a bounded
operator both in X = C[0, 1] (max norm) and X = L
2
cont
(0, 1).
Problem 0.10. Show that the diﬀerential operator A =
d
dx
deﬁned on
D(A) = C
1
[0, 1] ⊂ C[0, 1] is an unbounded operator.
22 0. A ﬁrst look at Banach and Hilbert spaces
Problem 0.11. Show that AB ≤ AB for every A, B ∈ L(X).
Problem 0.12. Show that the multiplication in a Banach algebra X is con
tinuous: x
n
→ x and y
n
→ y implies x
n
y
n
→ xy.
0.6. Lebesgue L
p
spaces
We ﬁx some measure space (X, Σ, µ) and deﬁne the L
p
norm by
f
p
=
X
[f[
p
dµ
1/p
, 1 ≤ p (0.66)
and denote by L
p
(X, dµ) the set of all complex valued measurable functions
for which f
p
is ﬁnite. First of all note that L
p
(X, dµ) is a linear space,
since [f + g[
p
≤ 2
p
max([f[, [g[)
p
≤ 2
p
max([f[
p
, [g[
p
) ≤ 2
p
([f[
p
+ [g[
p
). Of
course our hope is that L
p
(X, dµ) is a Banach space. However, there is
a small technical problem (recall that a property is said to hold almost
everywhere if the set where it fails to hold is contained in a set of measure
zero):
Lemma 0.25. Let f be measurable, then
X
[f[
p
dµ = 0 (0.67)
if and only if f(x) = 0 almost everywhere with respect to µ.
Proof. Observe that we have A = ¦x[f(x) = 0¦ =
¸
n
A
n
, where A
n
=
¦x[ [f(x)[ >
1
n
¦. If
[f[
p
dµ = 0 we must have µ(A
n
) = 0 for every n and
hence µ(A) = lim
n→∞
µ(A
n
) = 0. The converse is obvious.
Note that the proof also shows that if f is not 0 almost everywhere,
there is an ε > 0 such that µ(¦x[ [f(x)[ ≥ ε¦) > 0.
Example. Let λ be the Lebesgue measure on R. Then the characteristic
function of the rationals χ
Q
is zero a.e. (with respect to λ). Let Θ be the
Dirac measure centered at 0, then f(x) = 0 a.e. (with respect to Θ) if and
only if f(0) = 0.
Thus f
p
= 0 only implies f(x) = 0 for almost every x, but not for all!
Hence .
p
is not a norm on L
p
(X, dµ). The way out of this misery is to
identify functions which are equal almost everywhere: Let
^(X, dµ) = ¦f[f(x) = 0 µalmost everywhere¦. (0.68)
Then ^(X, dµ) is a linear subspace of L
p
(X, dµ) and we can consider the
quotient space
L
p
(X, dµ) = L
p
(X, dµ)/^(X, dµ). (0.69)
0.6. Lebesgue L
p
spaces 23
If dµ is the Lebesgue measure on X ⊆ R
n
we simply write L
p
(X). Observe
that f
p
is well deﬁned on L
p
(X, dµ).
Even though the elements of L
p
(X, dµ) are strictly speaking equivalence
classes of functions, we will still call them functions for notational conve
nience. However, note that for f ∈ L
p
(X, dµ) the value f(x) is not well
deﬁned (unless there is a continuous representative and diﬀerent continuous
functions are in diﬀerent equivalence classes, e.g., in the case of Lebesgue
measure).
With this modiﬁcation we are back in business since L
p
(X, dµ) turns
out to be a Banach space. We will show this in the following sections.
But before that let us also deﬁne L
∞
(X, dµ). It should be the set of
bounded measurable functions B(X) together with the sup norm. The only
problem is that if we want to identify functions equal almost everywhere, the
supremum is no longer independent of the equivalence class. The solution
is the essential supremum
f
∞
= inf¦C [ µ(¦x[ [f(x)[ > C¦) = 0¦. (0.70)
That is, C is an essential bound if [f(x)[ ≤ C almost everywhere and the
essential supremum is the inﬁmum over all essential bounds.
Example. If λ is the Lebesgue measure, then the essential sup of χ
Q
with
respect to λ is 0. If Θ is the Dirac measure centered at 0, then the essential
sup of χ
Q
with respect to Θ is 1 (since χ
Q
(0) = 1, and x = 0 is the only
point which counts for Θ).
As before we set
L
∞
(X, dµ) = B(X)/^(X, dµ) (0.71)
and observe that f
∞
is independent of the equivalence class.
If you wonder where the ∞ comes from, have a look at Problem 0.13.
As a preparation for proving that L
p
is a Banach space, we will need
H¨older’s inequality, which plays a central role in the theory of L
p
spaces.
In particular, it will imply Minkowski’s inequality, which is just the triangle
inequality for L
p
.
Theorem 0.26 (H¨older’s inequality). Let p and q be dual indices, that is,
1
p
+
1
q
= 1 (0.72)
with 1 ≤ p ≤ ∞. If f ∈ L
p
(X, dµ) and g ∈ L
q
(X, dµ) then fg ∈ L
1
(X, dµ)
and
f g
1
≤ f
p
g
q
. (0.73)
24 0. A ﬁrst look at Banach and Hilbert spaces
Proof. The case p = 1, q = ∞ (respectively p = ∞, q = 1) follows directly
from the properties of the integral and hence it remains to consider 1 <
p, q < ∞.
First of all it is no restriction to assume f
p
= g
q
= 1. Then, using
the elementary inequality (Problem 0.14)
a
1/p
b
1/q
≤
1
p
a +
1
q
b, a, b ≥ 0, (0.74)
with a = [f[
p
and b = [g[
q
and integrating over X gives
X
[f g[dµ ≤
1
p
X
[f[
p
dµ +
1
q
X
[g[
q
dµ = 1 (0.75)
and ﬁnishes the proof.
As a consequence we also get
Theorem 0.27 (Minkowski’s inequality). Let f, g ∈ L
p
(X, dµ), then
f +g
p
≤ f
p
+g
p
. (0.76)
Proof. Since the cases p = 1, ∞ are straightforward, we only consider 1 <
p < ∞. Using [f +g[
p
≤ [f[ [f +g[
p−1
+[g[ [f +g[
p−1
we obtain from H¨older’s
inequality (note (p −1)q = p)
f +g
p
p
≤ f
p
(f +g)
p−1

q
+g
p
(f +g)
p−1

q
= (f
p
+g
p
)(f +g)
p−1
p
. (0.77)
This shows that L
p
(X, dµ) is a normed linear space. Finally it remains
to show that L
p
(X, dµ) is complete.
Theorem 0.28. The space L
p
(X, dµ) is a Banach space.
Proof. Suppose f
n
is a Cauchy sequence. It suﬃces to show that some
subsequence converges (show this). Hence we can drop some terms such
that
f
n+1
−f
n

p
≤
1
2
n
. (0.78)
Now consider g
n
= f
n
−f
n−1
(set f
0
= 0). Then
G(x) =
∞
¸
k=1
[g
k
(x)[ (0.79)
is in L
p
. This follows from
n
¸
k=1
[g
k
[
p
≤
n
¸
k=1
g
k
(x)
p
≤ f
1

p
+
1
2
(0.80)
0.6. Lebesgue L
p
spaces 25
using the monotone convergence theorem. In particular, G(x) < ∞ almost
everywhere and the sum
∞
¸
n=1
g
n
(x) = lim
n→∞
f
n
(x) (0.81)
is absolutely convergent for those x. Now let f(x) be this limit. Since
[f(x) − f
n
(x)[
p
converges to zero almost everywhere and [f(x) − f
n
(x)[
p
≤
2
p
G(x)
p
∈ L
1
, dominated convergence shows f −f
n

p
→ 0.
In particular, in the proof of the last theorem we have seen:
Corollary 0.29. If f
n
− f
p
→ 0 then there is a subsequence which con
verges pointwise almost everywhere.
It even turns out that L
p
is separable.
Lemma 0.30. Suppose X is a second countable topological space (i.e., it
has a countable basis) and µ is a regular Borel measure. Then L
p
(X, dµ),
1 ≤ p < ∞ is separable.
Proof. The set of all characteristic functions χ
A
(x) with A ∈ Σ and µ(A) <
∞, is total by construction of the integral. Now our strategy is as follows:
Using outer regularity we can restrict A to open sets and using the existence
of a countable base, we can restrict A to open sets from this base.
Fix A. By outer regularity, there is a decreasing sequence of open sets
O
n
such that µ(O
n
) → µ(A). Since µ(A) < ∞ it is no restriction to assume
µ(O
n
) < ∞, and thus µ(O
n
`A) = µ(O
n
) − µ(A) → 0. Now dominated
convergence implies χ
A
− χ
On

p
→ 0. Thus the set of all characteristic
functions χ
O
(x) with O open and µ(O) < ∞, is total. Finally let B be a
countable basis for the topology. Then, every open set O can be written as
O =
¸
∞
j=1
˜
O
j
with
˜
O
j
∈ B. Moreover, by considering the set of all ﬁnite
unions of elements from B it is no restriction to assume
¸
n
j=1
˜
O
j
∈ B. Hence
there is an increasing sequence
˜
O
n
O with
˜
O
n
∈ B. By monotone con
vergence, χ
O
−χ
˜
On

p
→ 0 and hence the set of all characteristic functions
χ
˜
O
with
˜
O ∈ B is total.
To ﬁnish this chapter, let us show that continuous functions are dense
in L
p
.
Theorem 0.31. Let X be a locally compact metric space and let µ be a
σﬁnite regular Borel measure. Then the set C
c
(X) of continuous functions
with compact support is dense in L
p
(X, dµ), 1 ≤ p < ∞.
Proof. As in the previous proof the set of all characteristic functions χ
K
(x)
with K compact is total (using inner regularity). Hence it suﬃces to show
26 0. A ﬁrst look at Banach and Hilbert spaces
that χ
K
(x) can be approximated by continuous functions. By outer regu
larity there is an open set O ⊃ K such that µ(O`K) ≤ ε. By Urysohn’s
lemma (Lemma 0.11) there is a continuous function f
ε
which is one on K
and 0 outside O. Since
X
[χ
K
−f
ε
[
p
dµ =
O\K
[f
ε
[
p
dµ ≤ µ(O`K) ≤ ε (0.82)
we have f
ε
−χ
K
 → 0 and we are done.
If X is some subset of R
n
we can do even better.
A nonnegative function u ∈ C
∞
c
(R
n
) is called a molliﬁer if
R
n
u(x)dx = 1 (0.83)
The standard molliﬁer is u(x) = exp(
1
x
2
−1
) for [x[ < 1 and u(x) = 0 else.
If we scale a molliﬁer according to u
k
(x) = k
n
u(k x) such that its mass is
preserved (u
k

1
= 1) and it concentrates more and more around the origin
E
T
u
k
we have the following result (Problem 0.17):
Lemma 0.32. Let u be a molliﬁer in R
n
and set u
k
(x) = k
n
u(k x). Then
for any (uniformly) continuous function f : R
n
→C we have that
f
k
(x) =
R
n
u
k
(x −y)f(y)dy (0.84)
is in C
∞
(R
n
) and converges to f (uniformly).
Now we are ready to prove
Theorem 0.33. If X ⊆ R
n
and µ is a Borel measure, then the set C
∞
c
(X) of
all smooth functions with compact support is dense in L
p
(X, dµ), 1 ≤ p < ∞.
Proof. By our previous result it suﬃces to show that any continuous func
tion f(x) with compact support can be approximated by smooth ones. By
setting f(x) = 0 for x ∈ X, it is no restriction to assume X = R
n
. Now
choose a molliﬁer u and observe that f
k
has compact support (since f
0.7. Appendix: The uniform boundedness principle 27
has). Moreover, since f has compact support it is uniformly continuous
and f
k
→ f uniformly. But this implies f
k
→ f in L
p
.
Problem 0.13. Suppose µ(X) < ∞. Show that
lim
p→∞
f
p
= f
∞
(0.85)
for any bounded measurable function.
Problem 0.14. Prove (0.74). (Hint: Show that f(x) = (1 − t) + tx − x
t
,
x > 0, 0 < t < 1 satisﬁes f(a/b) ≥ 0 = f(1).)
Problem 0.15. Show the following generalization of H¨older’s inequality
f g
r
≤ f
p
g
q
, (0.86)
where
1
p
+
1
q
=
1
r
.
Problem 0.16 (Lyapunov inequality). Let 0 < θ < 1. Show that if f ∈
L
p
1
∩ L
p
2
, then f ∈ L
p
and
f
p
≤ f
θ
p
1
f
1−θ
p
2
, (0.87)
where
1
p
=
θ
p
1
+
1−θ
p
2
.
Problem 0.17. Prove Lemma 0.32. (Hint: To show that f
k
is smooth use
Problem A.7 and A.8.)
Problem 0.18. Construct a function f ∈ L
p
(0, 1) which has a pole at every
rational number in [0, 1]. (Hint: Start with the function f
0
(x) = [x[
−α
which
has a single pole at 0, then f
j
(x) = f
0
(x −x
j
) has a pole at x
j
.)
0.7. Appendix: The uniform boundedness
principle
Recall that the interior of a set is the largest open subset (that is, the union
of all open subsets). A set is called nowhere dense if its closure has empty
interior. The key to several important theorems about Banach spaces is the
observation that a Banach space cannot be the countable union of nowhere
dense sets.
Theorem 0.34 (Baire category theorem). Let X be a complete metric space,
then X cannot be the countable union of nowhere dense sets.
Proof. Suppose X =
¸
∞
n=1
X
n
. We can assume that the sets X
n
are closed
and none of them contains a ball, that is, X`X
n
is open and nonempty for
every n. We will construct a Cauchy sequence x
n
which stays away from all
X
n
.
28 0. A ﬁrst look at Banach and Hilbert spaces
Since X`X
1
is open and nonempty there is a closed ball B
r
1
(x
1
) ⊆
X`X
1
. Reducing r
1
a little, we can even assume B
r
1
(x
1
) ⊆ X`X
1
. More
over, since X
2
cannot contain B
r
1
(x
1
) there is some x
2
∈ B
r
1
(x
1
) that is
not in X
2
. Since B
r
1
(x
1
) ∩ (X`X
2
) is open there is a closed ball B
r
2
(x
2
) ⊆
B
r
1
(x
1
) ∩ (X`X
2
). Proceeding by induction we obtain a sequence of balls
such that
B
rn
(x
n
) ⊆ B
r
n−1
(x
n−1
) ∩ (X`X
n
). (0.88)
Now observe that in every step we can choose r
n
as small as we please, hence
without loss of generality r
n
→ 0. Since by construction x
n
∈ B
r
N
(x
N
) for
n ≥ N, we conclude that x
n
is Cauchy and converges to some point x ∈ X.
But x ∈ B
rn
(x
n
) ⊆ X`X
n
for every n, contradicting our assumption that
the X
n
cover X.
(Sets which can be written as countable union of nowhere dense sets are
called of ﬁrst category. All other sets are second category. Hence the name
category theorem.)
In other words, if X
n
⊆ X is a sequence of closed subsets which cover
X, at least one X
n
contains a ball of radius ε > 0.
Now we come to the ﬁrst important consequence, the uniform bound
edness principle.
Theorem 0.35 (BanachSteinhaus). Let X be a Banach space and Y some
normed linear space. Let ¦A
α
¦ ⊆ L(X, Y ) be a family of bounded operators.
Suppose A
α
x ≤ C(x) is bounded for ﬁxed x ∈ X, then A
α
 ≤ C is
uniformly bounded.
Proof. Let
X
n
= ¦x[ A
α
x ≤ n for all α¦ =
¸
α
¦x[ A
α
x ≤ n¦, (0.89)
then
¸
n
X
n
= X by assumption. Moreover, by continuity of A
α
and the
norm, each X
n
is an intersection of closed sets and hence closed. By Baire’s
theorem at least one contains a ball of positive radius: B
ε
(x
0
) ⊂ X
n
. Now
observe
A
α
y ≤ A
α
(y +x
0
) +A
α
x
0
 ≤ n +A
α
x
0
 (0.90)
for y < ε. Setting y = ε
x
x
we obtain
A
α
x ≤
n +C(x
0
)
ε
x (0.91)
for any x.
Part 1
Mathematical
Foundations of
Quantum Mechanics
Chapter 1
Hilbert spaces
The phase space in classical mechanics is the Euclidean space R
2n
(for the n
position and n momentum coordinates). In quantum mechanics the phase
space is always a Hilbert space H. Hence the geometry of Hilbert spaces
stands at the outset of our investigations.
1.1. Hilbert spaces
Suppose H is a vector space. A map '., ..` : H H → C is called skew
linear form if it is conjugate linear in the ﬁrst and linear in the second
argument. A positive deﬁnite skew linear form is called inner product or
scalar product. Associated with every scalar product is a norm
ψ =
'ψ, ψ`. (1.1)
The triangle inequality follows from the CauchySchwarzBunjakowski
inequality:
['ψ, ϕ`[ ≤ ψ ϕ (1.2)
with equality if and only if ψ and ϕ are parallel.
If H is complete with respect to the above norm, it is called a Hilbert
space. It is no restriction to assume that H is complete since one can easily
replace it by its completion.
Example. The space L
2
(M, dµ) is a Hilbert space with scalar product given
by
'f, g` =
M
f(x)
∗
g(x)dµ(x). (1.3)
31
32 1. Hilbert spaces
Similarly, the set of all square summable sequences
2
(N) is a Hilbert space
with scalar product
'f, g` =
¸
j∈N
f
∗
j
g
j
. (1.4)
(Note that the second example is a special case of the ﬁrst one; take M = R
and µ a sum of Dirac measures.)
A vector ψ ∈ H is called normalized or unit vector if ψ = 1. Two
vectors ψ, ϕ ∈ H are called orthogonal or perpendicular (ψ ⊥ ϕ) if
'ψ, ϕ` = 0 and parallel if one is a multiple of the other. If ψ and ϕ are
orthogonal we have the Pythagorean theorem:
ψ +ϕ
2
= ψ
2
+ϕ
2
, ψ ⊥ ϕ, (1.5)
which is straightforward to check.
Suppose ϕ is a unit vector, then the projection of ψ in the direction of
ϕ is given by
ψ
= 'ϕ, ψ`ϕ (1.6)
and ψ
⊥
deﬁned via
ψ
⊥
= ψ −'ϕ, ψ`ϕ (1.7)
is perpendicular to ϕ.
These results can also be generalized to more than one vector. A set
of vectors ¦ϕ
j
¦ is called orthonormal set if 'ϕ
j
, ϕ
k
` = 0 for j = k and
'ϕ
j
, ϕ
j
` = 1.
Lemma 1.1. Suppose ¦ϕ
j
¦
n
j=0
is an orthonormal set. Then every ψ ∈ H
can be written as
ψ = ψ
+ψ
⊥
, ψ
=
n
¸
j=0
'ϕ
j
, ψ`ϕ
j
, (1.8)
where ψ
and ψ
⊥
are orthogonal. Moreover, 'ϕ
j
, ψ
⊥
` = 0 for all 1 ≤ j ≤ n.
In particular,
ψ
2
=
n
¸
j=0
['ϕ
j
, ψ`[
2
+ψ
⊥

2
. (1.9)
Moreover, every
ˆ
ψ in the span of ¦ϕ
j
¦
n
j=0
satisﬁes
ψ −
ˆ
ψ ≥ ψ
⊥
 (1.10)
with equality holding if and only if
ˆ
ψ = ψ
. In other words, ψ
is uniquely
characterized as the vector in the span of ¦ϕ
j
¦
n
j=0
being closest to ψ.
1.2. Orthonormal bases 33
Proof. A straightforward calculation shows 'ϕ
j
, ψ −ψ
` = 0 and hence ψ
and ψ
⊥
= ψ − ψ
are orthogonal. The formula for the norm follows by
applying (1.5) iteratively.
Now, ﬁx a vector
ˆ
ψ =
n
¸
j=0
c
j
ϕ
j
. (1.11)
in the span of ¦ϕ
j
¦
n
j=0
. Then one computes
ψ −
ˆ
ψ
2
= ψ
+ψ
⊥
−
ˆ
ψ
2
= ψ
⊥

2
+ψ
−
ˆ
ψ
2
= ψ
⊥

2
+
n
¸
j=0
[c
j
−'ϕ
j
, ψ`[
2
(1.12)
from which the last claim follows.
From (1.9) we obtain Bessel’s inequality
n
¸
j=0
['ϕ
j
, ψ`[
2
≤ ψ
2
(1.13)
with equality holding if and only if ψ lies in the span of ¦ϕ
j
¦
n
j=0
.
Recall that a scalar product can be recovered from its norm by virtue of
the polarization identity
'ϕ, ψ` =
1
4
ϕ +ψ
2
−ϕ −ψ
2
+ iϕ −iψ
2
−iϕ + iψ
2
. (1.14)
A bijective operator U ∈ L(H
1
, H
2
) is called unitary if U preserves
scalar products:
'Uϕ, Uψ`
2
= 'ϕ, ψ`
1
, ϕ, ψ ∈ H
1
. (1.15)
By the polarization identity this is the case if and only if U preserves norms:
Uψ
2
= ψ
1
for all ψ ∈ H
1
. The two Hilbert space H
1
and H
2
are called
unitarily equivalent in this case.
Problem 1.1. The operator S :
2
(N) →
2
(N), (a
1
, a
2
, a
3
. . . ) → (0, a
1
, a
2
, . . . )
clearly satisﬁes Ua = a. Is it unitary?
1.2. Orthonormal bases
Of course, since we cannot assume H to be a ﬁnite dimensional vector space,
we need to generalize Lemma 1.1 to arbitrary orthonormal sets ¦ϕ
j
¦
j∈J
.
We start by assuming that J is countable. Then Bessel’s inequality (1.13)
shows that
¸
j∈J
['ϕ
j
, ψ`[
2
(1.16)
34 1. Hilbert spaces
converges absolutely. Moreover, for any ﬁnite subset K ⊂ J we have

¸
j∈K
'ϕ
j
, ψ`ϕ
j

2
=
¸
j∈K
['ϕ
j
, ψ`[
2
(1.17)
by the Pythagorean theorem and thus
¸
j∈J
'ϕ
j
, ψ`ϕ
j
is Cauchy if and only
if
¸
j∈J
['ϕ
j
, ψ`[
2
is. Now let J be arbitrary. Again, Bessel’s inequality
shows that for any given ε > 0 there are at most ﬁnitely many j for which
['ϕ
j
, ψ`[ ≥ ε. Hence there are at most countably many j for which ['ϕ
j
, ψ`[ >
0. Thus it follows that
¸
j∈J
['ϕ
j
, ψ`[
2
(1.18)
is welldeﬁned and so is
¸
j∈J
'ϕ
j
, ψ`ϕ
j
. (1.19)
In particular, by continuity of the scalar product we see that Lemma 1.1
holds for arbitrary orthonormal sets without modiﬁcations.
Theorem 1.2. Suppose ¦ϕ
j
¦
j∈J
is an orthonormal set. Then every ψ ∈ H
can be written as
ψ = ψ
+ψ
⊥
, ψ
=
¸
j∈J
'ϕ
j
, ψ`ϕ
j
, (1.20)
where ψ
and ψ
⊥
are orthogonal. Moreover, 'ϕ
j
, ψ
⊥
` = 0 for all j ∈ J. In
particular,
ψ
2
=
¸
j∈J
['ϕ
j
, ψ`[
2
+ψ
⊥

2
. (1.21)
Moreover, every
ˆ
ψ in the span of ¦ϕ
j
¦
j∈J
satisﬁes
ψ −
ˆ
ψ ≥ ψ
⊥
 (1.22)
with equality holding if and only if
ˆ
ψ = ψ
. In other words, ψ
is uniquely
characterized as the vector in the span of ¦ϕ
j
¦
j∈J
being closest to ψ.
Note that from Bessel’s inequality (which of course still holds) it follows
that the map ψ → ψ
is continuous.
An orthonormal set which is not a proper subset of any other orthonor
mal set is called an orthonormal basis due to following result:
Theorem 1.3. For an orthonormal set ¦ϕ
j
¦
j∈J
the following conditions are
equivalent:
(i) ¦ϕ
j
¦
j∈J
is a maximal orthonormal set.
(ii) For every vector ψ ∈ H we have
ψ =
¸
j∈J
'ϕ
j
, ψ`ϕ
j
. (1.23)
1.2. Orthonormal bases 35
(iii) For every vector ψ ∈ H we have
ψ
2
=
¸
j∈J
['ϕ
j
, ψ`[
2
. (1.24)
(iv) 'ϕ
j
, ψ` = 0 for all j ∈ J implies ψ = 0.
Proof. We will use the notation from Theorem 1.2.
(i) ⇒ (ii): If ψ
⊥
= 0 than we can normalize ψ
⊥
to obtain a unit vector
˜
ψ
⊥
which is orthogonal to all vectors ϕ
j
. But then ¦ϕ
j
¦
j∈J
∪ ¦
˜
ψ
⊥
¦ would be a
larger orthonormal set, contradicting maximality of ¦ϕ
j
¦
j∈J
.
(ii) ⇒ (iii): Follows since (ii) implies ψ
⊥
= 0.
(iii) ⇒ (iv): If 'ψ, ϕ
j
` = 0 for all j ∈ J we conclude ψ
2
= 0 and hence
ψ = 0.
(iv) ⇒ (i): If ¦ϕ
j
¦
j∈J
were not maximal, there would be a unit vector ϕ
such that ¦ϕ
j
¦
j∈J
∪ ¦ϕ¦ is larger orthonormal set. But 'ϕ
j
, ϕ` = 0 for all
j ∈ J implies ϕ = 0 by (iv), a contradiction.
Since ψ → ψ
is continuous, it suﬃces to check conditions (ii) and (iii)
on a dense set.
Example. The set of functions
ϕ
n
(x) =
1
√
2π
e
inx
, n ∈ Z, (1.25)
forms an orthonormal basis for H = L
2
(0, 2π). The corresponding orthogo
nal expansion is just the ordinary Fourier series. (Problem 1.17)
A Hilbert space is separable if and only if there is a countable orthonor
mal basis. In fact, if H is separable, then there exists a countable total set
¦ψ
j
¦
N
j=0
. After throwing away some vectors we can assume that ψ
n+1
can
not be expressed as a linear combinations of the vectors ψ
0
, . . . ψ
n
. Now we
can construct an orthonormal basis as follows: We begin by normalizing ψ
0
ϕ
0
=
ψ
0
ψ
0

. (1.26)
Next we take ψ
1
and remove the component parallel to ϕ
0
and normalize
again
ϕ
1
=
ψ
1
−'ϕ
0
, ψ
1
`ϕ
0
ψ
1
−'ϕ
0
, ψ
1
`ϕ
0

. (1.27)
Proceeding like this we deﬁne recursively
ϕ
n
=
ψ
n
−
¸
n−1
j=0
'ϕ
j
, ψ
n
`ϕ
j
ψ
n
−
¸
n−1
j=0
'ϕ
j
, ψ
n
`ϕ
j

. (1.28)
This procedure is known as GramSchmidt orthogonalization. Hence
we obtain an orthonormal set ¦ϕ
j
¦
N
j=0
such that span¦ϕ
j
¦
n
j=0
= span¦ψ
j
¦
n
j=0
36 1. Hilbert spaces
for any ﬁnite n and thus also for N. Since ¦ψ
j
¦
N
j=0
is total, we infer that
¦ϕ
j
¦
N
j=0
is an orthonormal basis.
Example. In L
2
(−1, 1) we can orthogonalize the polynomial f
n
(x) = x
n
.
The resulting polynomials are up to a normalization equal to the Legendre
polynomials
P
0
(x) = 1, P
1
(x) = x, P
2
(x) =
3 x
2
−1
2
, . . . (1.29)
(which are normalized such that P
n
(1) = 1).
If fact, if there is one countable basis, then it follows that any other basis
is countable as well.
Theorem 1.4. If H is separable, then every orthonormal basis is countable.
Proof. We know that there is at least one countable orthonormal basis
¦ϕ
j
¦
j∈J
. Now let ¦φ
k
¦
k∈K
be a second basis and consider the set K
j
=
¦k ∈ K['φ
k
, ϕ
j
` = 0¦. Since these are the expansion coeﬃcients of ϕ
j
with
respect to ¦φ
k
¦
k∈K
, this set is countable. Hence the set
˜
K =
¸
j∈J
K
j
is
countable as well. But k ∈ K`
˜
K implies φ
k
= 0 and hence
˜
K = K.
We will assume all Hilbert spaces to be separable.
In particular, it can be shown that L
2
(M, dµ) is separable. Moreover, it
turns out that, up to unitary equivalence, there is only one (separable)
inﬁnite dimensional Hilbert space:
Let H be an inﬁnite dimensional Hilbert space and let ¦ϕ
j
¦
j∈N
be any
orthogonal basis. Then the map U : H →
2
(N), ψ → ('ϕ
j
, ψ`)
j∈N
is unitary
(by Theorem 1.3 (iii)). In particular,
Theorem 1.5. Any separable inﬁnite dimensional Hilbert space is unitarily
equivalent to
2
(N).
Let me remark that if H is not separable, there still exists an orthonor
mal basis. However, the proof requires Zorn’s lemma: The collection of
all orthonormal sets in H can be partially ordered by inclusion. Moreover,
any linearly ordered chain has an upper bound (the union of all sets in the
chain). Hence Zorn’s lemma implies the existence of a maximal element,
that is, an orthonormal basis.
1.3. The projection theorem and the Riesz
lemma
Let M ⊆ H be a subset, then M
⊥
= ¦ψ['ϕ, ψ` = 0, ∀ϕ ∈ M¦ is called
the orthogonal complement of M. By continuity of the scalar prod
uct it follows that M
⊥
is a closed linear subspace and by linearity that
1.3. The projection theorem and the Riesz lemma 37
(span(M))
⊥
= M
⊥
. For example we have H
⊥
= ¦0¦ since any vector in H
⊥
must be in particular orthogonal to all vectors in some orthonormal basis.
Theorem 1.6 (projection theorem). Let M be a closed linear subspace of a
Hilbert space H, then every ψ ∈ H can be uniquely written as ψ = ψ
+ ψ
⊥
with ψ
∈ M and ψ
⊥
∈ M
⊥
. One writes
M ⊕M
⊥
= H (1.30)
in this situation.
Proof. Since M is closed, it is a Hilbert space and has an orthonormal basis
¦ϕ
j
¦
j∈J
. Hence the result follows from Theorem 1.2.
In other words, to every ψ ∈ H we can assign a unique vector ψ
which
is the vector in M closest to ψ. The rest ψ − ψ
lies in M
⊥
. The operator
P
M
ψ = ψ
is called the orthogonal projection corresponding to M. Note
that we have
P
2
M
= P
M
and 'P
M
ψ, ϕ` = 'ψ, P
M
ϕ` (1.31)
since 'P
M
ψ, ϕ` = 'ψ
, ϕ
` = 'ψ, P
M
ϕ`. Clearly we have P
M
⊥ψ = ψ −
P
M
ψ = ψ
⊥
.
Moreover, we see that the vectors in a closed subspace M are precisely
those which are orthogonal to all vectors in M
⊥
, that is, M
⊥⊥
= M. If M
is an arbitrary subset we have at least
M
⊥⊥
= span(M). (1.32)
Note that by H
⊥
= ¦¦ we see that M
⊥
= ¦0¦ if and only if M is dense.
Finally we turn to linear functionals, that is, to operators : H →
C. By the CauchySchwarz inequality we know that
ϕ
: ψ → 'ϕ, ψ` is a
bounded linear functional (with norm ϕ). In turns out that in a Hilbert
space every bounded linear functional can be written in this way.
Theorem 1.7 (Riesz lemma). Suppose is a bounded linear functional on
a Hilbert space H. Then there is a vector ϕ ∈ H such that (ψ) = 'ϕ, ψ` for
all ψ ∈ H. In other words, a Hilbert space is equivalent to its own dual space
H
∗
= H.
Proof. If ≡ 0 we can choose ϕ = 0. Otherwise Ker() = ¦ψ[(ψ) = 0¦
is a proper subspace and we can ﬁnd a unit vector ˜ ϕ ∈ Ker()
⊥
. For every
ψ ∈ H we have (ψ) ˜ ϕ −( ˜ ϕ)ψ ∈ Ker() and hence
0 = ' ˜ ϕ, (ψ) ˜ ϕ −( ˜ ϕ)ψ` = (ψ) −( ˜ ϕ)' ˜ ϕ, ψ`. (1.33)
In other words, we can choose ϕ = ( ˜ ϕ)
∗
˜ ϕ.
The following easy consequence is left as an exercise.
38 1. Hilbert spaces
Corollary 1.8. Suppose B is a bounded skew liner form, that is,
[B(ψ, ϕ)[ ≤ Cψ ϕ. (1.34)
Then there is a unique bounded operator A such that
B(ψ, ϕ) = 'Aψ, ϕ`. (1.35)
Problem 1.2. Show that an orthogonal projection P
M
= 0 has norm one.
Problem 1.3. Suppose P
1
and P
1
are orthogonal projections. Show that
P
1
≤ P
2
(that is 'ψ, P
1
ψ` ≤ 'ψ, P
2
ψ`) is equivalent to Ran(P
1
) ⊆ Ran(P
2
).
Problem 1.4. Prove Corollary 1.8.
Problem 1.5. Let ¦ϕ
j
¦ be some orthonormal basis. Show that a bounded
linear operator A is uniquely determined by its matrix elements A
jk
=
'ϕ
j
, Aϕ
k
` with respect to this basis.
Problem 1.6. Show that L(H) is not separable H is inﬁnite dimensional.
Problem 1.7. Show P : L
2
(R) → L
2
(R), f(x) →
1
2
(f(x) + f(−x)) is a
projection. Compute its range and kernel.
1.4. Orthogonal sums and tensor products
Given two Hilbert spaces H
1
and H
2
we deﬁne their orthogonal sumH
1
⊕H
2
to be the set of all pairs (ψ
1
, ψ
2
) ∈ H
1
H
2
together with the scalar product
'(ϕ
1
, ϕ
2
), (ψ
1
, ψ
2
)` = 'ϕ
1
, ψ
1
`
1
+'ϕ
2
, ψ
2
`
2
. (1.36)
It is left as an exercise to verify that H
1
⊕ H
2
is again a Hilbert space.
Moreover, H
1
can be identiﬁed with ¦(ψ
1
, 0)[ψ
1
∈ H
1
¦ and we can regard
H
1
as a subspace of H
1
⊕ H
2
. Similarly for H
2
. It is also custom to write
ψ
1
+ψ
2
instead of (ψ
1
, ψ
2
).
More generally, let H
j
j ∈ N, be a countable collection of Hilbert spaces
and deﬁne
∞
j=1
H
j
= ¦
∞
¸
j=1
ψ
j
[ ψ
j
∈ H
j
,
∞
¸
j=1
ψ
j

2
j
< ∞¦, (1.37)
which becomes a Hilbert space with the scalar product
'
∞
¸
j=1
ϕ
j
,
∞
¸
j=1
ψ
j
` =
∞
¸
j=1
'ϕ
j
, ψ
j
`
j
. (1.38)
Suppose H and
˜
H are two Hilbert spaces. Our goal is to construct their
tensor product. The elements should be products ψ ⊗
˜
ψ of elements ψ ∈ H
1.4. Orthogonal sums and tensor products 39
and
˜
ψ ∈
˜
H. Hence we start with the set of all ﬁnite linear combinations of
elements of H
˜
H
T(H,
˜
H) = ¦
n
¸
j=1
α
j
(ψ
j
,
˜
ψ
j
)[(ψ
j
,
˜
ψ
j
) ∈ H
˜
H, α
j
∈ C¦. (1.39)
Since we want (ψ
1
+ψ
2
)⊗
˜
ψ = ψ
1
⊗
˜
ψ+ψ
2
⊗
˜
ψ, ψ⊗(
˜
ψ
1
+
˜
ψ
2
) = ψ⊗
˜
ψ
1
+ψ⊗
˜
ψ
2
,
and (αψ) ⊗
˜
ψ = ψ ⊗(α
˜
ψ) we consider T(H,
˜
H)/^(H,
˜
H), where
^(H,
˜
H) = span¦
n
¸
j,k=1
α
j
β
k
(ψ
j
,
˜
ψ
k
) −(
n
¸
j=1
α
j
ψ
j
,
n
¸
k=1
β
k
˜
ψ
k
)¦ (1.40)
and write ψ ⊗
˜
ψ for the equivalence class of (ψ,
˜
ψ).
Next we deﬁne
'ψ ⊗
˜
ψ, φ ⊗
˜
φ` = 'ψ, φ`'
˜
ψ,
˜
φ` (1.41)
which extends to a skew linear form on T(H,
˜
H)/^(H,
˜
H). To show that we
obtain a scalar product, we need to ensure positivity. Let ψ =
¸
i
α
i
ψ
i
⊗
˜
ψ
i
=
0 and pick orthonormal bases ϕ
j
, ˜ ϕ
k
for span¦ψ
i
¦, span¦
˜
ψ
i
¦, respectively.
Then
ψ =
¸
j,k
α
jk
ϕ
j
⊗ ˜ ϕ
k
, α
jk
=
¸
i
α
i
'ϕ
j
, ψ
i
`' ˜ ϕ
k
,
˜
ψ
i
` (1.42)
and we compute
'ψ, ψ` =
¸
j,k
[α
jk
[
2
> 0. (1.43)
The completion of T(H,
˜
H)/^(H,
˜
H) with respect to the induced norm is
called the tensor product H ⊗
˜
H of H and
˜
H.
Lemma 1.9. If ϕ
j
, ˜ ϕ
k
are orthonormal bases for H,
˜
H, respectively, then
ϕ
j
⊗ ˜ ϕ
k
is an orthonormal basis for H ⊗
˜
H.
Proof. That ϕ
j
⊗ ˜ ϕ
k
is an orthonormal set is immediate from (1.41). More
over, since span¦ϕ
j
¦, span¦ ˜ ϕ
k
¦ is dense in H,
˜
H, respectively, it is easy to
see that ϕ
j
⊗ ˜ ϕ
k
is dense in T(H,
˜
H)/^(H,
˜
H). But the latter is dense in
H ⊗
˜
H.
Example. We have H ⊗C
n
= H
n
.
Example. Let (M, dµ) and (
˜
M, d˜ µ) be two measure spaces. Then we have
L
2
(M, dµ) ⊗L
2
(
˜
M, d˜ µ) = L
2
(M
˜
M, dµ d˜ µ).
Clearly we have L
2
(M, dµ) ⊗ L
2
(
˜
M, d˜ µ) ⊆ L
2
(M
˜
M, dµ d˜ µ). Now
take an orthonormal basis ϕ
j
⊗ ˜ ϕ
k
for L
2
(M, dµ) ⊗ L
2
(
˜
M, d˜ µ) as in our
previous lemma. Then
M
˜
M
(ϕ
j
(x) ˜ ϕ
k
(y))
∗
f(x, y)dµ(x)d˜ µ(y) = 0 (1.44)
40 1. Hilbert spaces
implies
M
ϕ
j
(x)
∗
f
k
(x)dµ(x) = 0, f
k
(x) =
˜
M
˜ ϕ
k
(y)
∗
f(x, y)d˜ µ(y) (1.45)
and hence f
k
(x) = 0 µa.e. x. But this implies f(x, y) = 0 for µa.e. x and
˜ µa.e. y and thus f = 0. Hence ϕ
j
⊗ ˜ ϕ
k
is a basis for L
2
(M
˜
M, dµ d˜ µ)
and equality follows.
It is straightforward to extend the tensor product to any ﬁnite number
of Hilbert spaces. We even note
(
∞
j=1
H
j
) ⊗H =
∞
j=1
(H
j
⊗H), (1.46)
where equality has to be understood in the sense, that both spaces are
unitarily equivalent by virtue of the identiﬁcation
(
∞
¸
j=1
ψ
j
) ⊗ψ =
∞
¸
j=1
ψ
j
⊗ψ. (1.47)
Problem 1.8. We have ψ⊗
˜
ψ = φ⊗
˜
φ if and only if there is some α ∈ C`¦0¦
such that ψ = αφ and
˜
ψ = α
−1
˜
φ.
Problem 1.9. Show (1.46)
1.5. The C
∗
algebra of bounded linear operators
We start by introducing a conjugation for operators on a Hilbert space H.
Let A = L(H), then the adjoint operator is deﬁned via
'ϕ, A
∗
ψ` = 'Aϕ, ψ` (1.48)
(compare Corollary 1.8).
Example. If H = C
n
and A = (a
jk
)
1≤j,k≤n
, then A
∗
= (a
∗
kj
)
1≤j,k≤n
.
Lemma 1.10. Let A, B ∈ L(H), then
(i) (A+B)
∗
= A
∗
+B
∗
, (αA)
∗
= α
∗
A
∗
,
(ii) A
∗∗
= A,
(iii) (AB)
∗
= B
∗
A
∗
,
(iv) A = A
∗
 and A
2
= A
∗
A = AA
∗
.
Proof. (i) and (ii) are obvious. (iii) follows from'ϕ, (AB)ψ` = 'A
∗
ϕ, Bψ` =
'B
∗
A
∗
ϕ, ψ`. (iv) follows from
A
∗
 = sup
ϕ=ψ=1
['ψ, A
∗
ϕ`[ = sup
ϕ=ψ=1
['Aψ, ϕ`[ = A. (1.49)
1.6. Weak and strong convergence 41
and
A
∗
A = sup
ϕ=ψ=1
['ϕ, A
∗
Aψ`[ = sup
ϕ=ψ=1
['Aϕ, Aψ`[
= sup
ϕ=1
Aϕ
2
= A
2
, (1.50)
where we have used ϕ = sup
ψ=1
['ψ, ϕ`[.
As a consequence of A
∗
 = A observe that taking the adjoint is
continuous.
In general, a Banach algebra / together with an involution
(a +b)
∗
= a
∗
+b
∗
, (αa)
∗
= α
∗
a
∗
, a
∗∗
= a, (ab)
∗
= b
∗
a
∗
, (1.51)
satisfying
a
2
= a
∗
a (1.52)
is called a C
∗
algebra. The element a
∗
is called the adjoint of a. Note that
a
∗
 = a follows from (1.52) and aa
∗
 ≤ aa
∗
.
Any subalgebra which is also closed under involution, is called a ∗
algebra. An ideal is a subspace 1 ⊆ / such that a ∈ 1, b ∈ / implies
ab ∈ 1 and ba ∈ 1. If it is closed under the adjoint map it is called a ∗ideal.
Note that if there is and identity e we have e
∗
= e and hence (a
−1
)
∗
= (a
∗
)
−1
(show this).
Example. The continuous function C(I) together with complex conjuga
tion form a commutative C
∗
algebra.
An element a ∈ / is called normal if aa
∗
= a
∗
a, selfadjoint if a = a
∗
,
unitary if aa
∗
= a
∗
a = I, (orthogonal) projection if a = a
∗
= a
2
, and
positive if a = bb
∗
for some b ∈ /. Clearly both selfadjoint and unitary
elements are normal.
Problem 1.10. Let A ∈ L(H). Show that A is normal if and only if
Aψ = A
∗
ψ, ∀ψ ∈ H. (1.53)
(Hint: Problem 0.6)
Problem 1.11. Show that U : H →H is unitary if and only if U
−1
= U
∗
.
Problem 1.12. Compute the adjoint of S :
2
(N) →
2
(N), (a
1
, a
2
, a
3
. . . ) →
(0, a
1
, a
2
, . . . ).
1.6. Weak and strong convergence
Sometimes a weaker notion of convergence is useful: We say that ψ
n
con
verges weakly to ψ and write
wlim
n→∞
ψ
n
= ψ or ψ
n
ψ. (1.54)
42 1. Hilbert spaces
if 'ϕ, ψ
n
` → 'ϕ, ψ` for every ϕ ∈ H (show that a weak limit is unique).
Example. Let ϕ
n
be an (inﬁnite) orthonormal set. Then 'ψ, ϕ
n
` → 0 for
every ψ since these are just the expansion coeﬃcients of ψ. (ϕ
n
does not
converge to 0, since ϕ
n
 = 1.)
Clearly ψ
n
→ ψ implies ψ
n
ψ and hence this notion of convergence is
indeed weaker. Moreover, the weak limit is unique, since 'ϕ, ψ
n
` → 'ϕ, ψ`
and 'ϕ, ψ
n
` → 'ϕ,
˜
ψ` implies 'ϕ, (ψ−
˜
ψ)` = 0. A sequence ψ
n
is called weak
Cauchy sequence if 'ϕ, ψ
n
` is Cauchy for every ϕ ∈ H.
Lemma 1.11. Let H be a Hilbert space.
(i) ψ
n
ψ implies ψ ≤ liminf ψ
n
.
(ii) Every weak Cauchy sequence ψ
n
is bounded: ψ
n
 ≤ C.
(iii) Every weak Cauchy sequence converges weakly.
(iv) For a weakly convergent sequence ψ
n
ψ we have: ψ
n
→ ψ if
and only if limsup ψ
n
 ≤ ψ.
Proof. (i) Observe
ψ
2
= 'ψ, ψ` = liminf'ψ, ψ
n
` ≤ ψ liminf ψ
n
. (1.55)
(ii) For every ϕ we have that ['ϕ, ψ
n
`[ ≤ C(ϕ) is bounded. Hence by the
uniform boundedness principle we have ψ
n
 = 'ψ
n
, .` ≤ C.
(iii) Let ϕ
m
be an orthonormal basis and deﬁne c
m
= lim
n→∞
'ϕ
m
, ψ
n
`.
Then ψ =
¸
m
c
m
ϕ
m
is the desired limit.
(iv) By (i) we have limψ
n
 = ψ and hence
ψ −ψ
n

2
= ψ
2
−2Re('ψ, ψ
n
`) +ψ
n

2
→ 0. (1.56)
The converse is straightforward.
Clearly an orthonormal basis does not have a norm convergent subse
quence. Hence the unit ball in an inﬁnite dimensional Hilbert space is never
compact. However, we can at least extract weakly convergent subsequences:
Lemma 1.12. Let H be a Hilbert space. Every bounded sequence ψ
n
has
weakly convergent subsequence.
Proof. Let ϕ
k
be an ONB, then by the usual diagonal sequence argument
we can ﬁnd a subsequence ψ
nm
such that 'ϕ
k
, ψ
nm
` converges for all k. Since
ψ
n
is bounded, 'ϕ, ψ
nm
` converges for every ϕ ∈ H and hence ψ
nm
is a weak
Cauchy sequence.
Finally, let me remark that similar concepts can be introduced for oper
ators. This is of particular importance for the case of unbounded operators,
where convergence in the operator norm makes no sense at all.
1.6. Weak and strong convergence 43
A sequence of operators A
n
is said to converge strongly to A,
slim
n→∞
A
n
= A :⇔ A
n
ψ → Aψ ∀x ∈ D(A) ⊆ D(A
n
). (1.57)
It is said to converge weakly to A,
wlim
n→∞
A
n
= A :⇔ A
n
ψ Aψ ∀ψ ∈ D(A) ⊆ D(A
n
). (1.58)
Clearly norm convergence implies strong convergence and strong conver
gence implies weak convergence.
Example. Consider the operator S
n
∈ L(
2
(N)) which shifts a sequence n
places to the left
S
n
(x
1
, x
2
, . . . ) = (x
n+1
, x
n+2
, . . . ). (1.59)
and the operator S
∗
n
∈ L(
2
(N)) which shifts a sequence n places to the right
and ﬁlls up the ﬁrst n places with zeros
S
∗
n
(x
1
, x
2
, . . . ) = (0, . . . , 0
. .. .
n places
, x
1
, x
2
, . . . ). (1.60)
Then S
n
converges to zero strongly but not in norm and S
∗
n
converges weakly
to zero but not strongly.
Note that this example also shows that taking adjoints is not continuous
with respect to strong convergence! If A
n
s
→ A we only have
'ϕ, A
∗
n
ψ` = 'A
n
ϕ, ψ` → 'Aϕ, ψ` = 'ϕ, A
∗
ψ` (1.61)
and hence A
∗
n
A
∗
in general. However, if A
n
and A are normal we have
(A
n
−A)
∗
ψ = (A
n
−A)ψ (1.62)
and hence A
∗
n
s
→ A
∗
in this case. Thus at least for normal operators taking
adjoints is continuous with respect to strong convergence.
Lemma 1.13. Suppose A
n
is a sequence of bounded operators.
(i) slim
n→∞
A
n
= A implies A ≤ liminf A
n
.
(ii) Every strong Cauchy sequence A
n
is bounded: A
n
 ≤ C.
(iii) If A
n
y → Ay for y in a dense set and A
n
 ≤ C, than slim
n→∞
A
n
=
A.
The same result holds if strong convergence is replaced by weak convergence.
Proof. (i) and (ii) follow as in Lemma 1.11 (i).
(ii) Just use
A
n
ψ −Aψ ≤ A
n
ψ −A
n
ϕ +A
n
ϕ −Aϕ +Aϕ −Aψ
≤ 2Cψ −ϕ +A
n
ϕ −Aϕ (1.63)
44 1. Hilbert spaces
and choose ϕ in the dense subspace such that ψ − ϕ ≤
ε
4C
and n large
such that A
n
ϕ −Aϕ ≤
ε
2
.
The case of weak convergence is left as an exercise.
Problem 1.13. Suppose ψ
n
→ ψ and ϕ
n
ϕ. Then 'ψ
n
, ϕ
n
` → 'ψ, ϕ`.
Problem 1.14. Let ¦ϕ
j
¦ be some orthonormal basis. Show that ψ
n
ϕ
if and only if ψ
n
is bounded and 'ϕ
j
, ψ
n
` → 'ϕ
j
, ψ` for every j. Show that
this is wrong without the boundedness assumption.
Problem 1.15. Let ¦ϕ
j
¦
∞
j=1
be some orthonormal basis and deﬁne
ψ
w
=
∞
¸
j=1
1
2
j
['ϕ
j
, ψ`[. (1.64)
Show that .
w
is a norm. Show that ψ
n
ϕ if and only if ψ
n
−ψ
w
→ 0.
Problem 1.16. A subspace M ⊆ H is closed if and only if every weak
Cauchy sequence in M has a limit in M. (Hint: M = M
⊥⊥
.)
1.7. Appendix: The Stone–Weierstraß theorem
In case of a selfadjoint operator, the spectral theorem will show that the
closed ∗algebra generated by this operator is isomorphic to the C
∗
algebra
of continuous functions C(K) over some compact set. Hence it is important
to be able to identify dense sets:
Theorem 1.14 (Stone–Weierstraß, real version). Suppose K is a compact
set and let C(K, R) be the Banach algebra of continuous functions (with the
sup norm).
If F ⊂ C(K, R) contains the identity 1 and separates points (i.e., for
every x
1
= x
2
there is some function f ∈ F such that f(x
1
) = f(x
2
)), then
the algebra generated by F is dense.
Proof. Denote by A the algebra generated by F. Note that if f ∈ A,
we have [f[ ∈ A: By the Weierstraß approximation theorem (Theorem 0.13)
there is a polynomial p
n
(t) such that
[t[−p
n
(t)
<
1
n
and hence p
n
(f) → [f[.
In particular, if f, g in A, we also have
max¦f, g¦ =
(f +g) +[f −g[
2
, min¦f, g¦ =
(f +g) −[f −g[
2
(1.65)
in A.
Now ﬁx f ∈ C(K, R). We need to ﬁnd some f
ε
∈ A with f −f
ε

∞
< ε.
First of all, since A separates points, observe that for given y, z ∈ K
there is a function f
y,z
∈ A such that f
y,z
(y) = f(y) and f
y,z
(z) = f(z)
1.7. Appendix: The Stone–Weierstraß theorem 45
(show this). Next, for every y ∈ K there is a neighborhood U(y) such that
f
y,z
(x) > f(x) −ε, x ∈ U(y) (1.66)
and since K is compact, ﬁnitely many, say U(y
1
), . . . U(y
j
), cover K. Then
f
z
= max¦f
y
1
,z
, . . . , f
y
j
,z
¦ ∈ A (1.67)
and satisﬁes f
z
> f −ε by construction. Since f
z
(z) = f(z) for every z ∈ K
there is a neighborhood V (z) such that
f
z
(x) < f(x) +ε, x ∈ V (z) (1.68)
and a corresponding ﬁnite cover V (z
1
), . . . V (z
k
). Now
f
ε
= min¦f
z
1
, . . . , f
z
k
¦ ∈ A (1.69)
satisﬁes f
ε
< f + ε. Since f − ε < f
z
l
< f
ε
, we have found a required
function.
Theorem 1.15 (Stone–Weierstraß). Suppose K is a compact set and let
C(K) be the C
∗
algebra of continuous functions (with the sup norm).
If F ⊂ C(K) contains the identity 1 and separates points, then the ∗
algebra generated by F is dense.
Proof. Just observe that
˜
F = ¦Re(f), Im(f)[f ∈ F¦ satisﬁes the assump
tion of the real version. Hence any realvalued continuous functions can be
approximated by elements from
˜
F, in particular this holds for the real and
imaginary part for any given complexvalued function.
Note that the additional requirement of being closed under complex
conjugation is crucial: The functions holomorphic on the unit ball and con
tinuous on the boundary separate points, but they are not dense (since the
uniform limit of holomorphic functions is again holomorphic).
Corollary 1.16. Suppose K is a compact set and let C(K) be the C
∗
algebra
of continuous functions (with the sup norm).
If F ⊂ C(K) separates points, then the closure of the ∗algebra generated
by F is either C(K) or ¦f ∈ C(K)[f(t
0
) = 0¦ for some t
0
∈ K.
Proof. There are two possibilities, either all f ∈ F vanish at one point
t
0
∈ K (there can be at most one such point since F separates points)
or there is no such point. If there is no such point we can proceed as in
the proof of the Stone–Weierstraß theorem to show that the identity can
be approximated by elements in A (note that to show [f[ ∈ A if f ∈ A
we do not need the identity, since p
n
can be chosen to contain no constant
term). If there is such a t
0
, the identity is clearly missing from A. However,
adding the identity to A we get A + C = C(K) and it is easy to see that
A = ¦f ∈ C(K)[f(t
0
) = 0¦.
46 1. Hilbert spaces
Problem 1.17. Show that the of functions ϕ
n
(x) =
1
√
2π
e
inx
, n ∈ Z, form
an orthonormal basis for H = L
2
(0, 2π).
Problem 1.18. Show that the ∗algebra generated by f
z
(t) =
1
t−z
, z ∈ C, is
dense in the C
∗
algebra C
∞
(R) of continuous functions vanishing at inﬁnity.
(Hint: Add ∞ to R to make it compact.)
Chapter 2
Selfadjointness and
spectrum
2.1. Some quantum mechanics
In quantum mechanics, a single particle living in R
3
is described by a
complexvalued function (the wave function)
ψ(x, t), (x, t) ∈ R
3
R, (2.1)
where x corresponds to a point in space and t corresponds to time. The
quantity ρ
t
(x) = [ψ(x, t)[
2
is interpreted as the probability density of the
particle at the time t. In particular, ψ must be normalized according to
R
3
[ψ(x, t)[
2
d
3
x = 1, t ∈ R. (2.2)
The location x of the particle is a quantity which can be observed (i.e.,
measured) and is hence called observable. Due to our probabilistic inter
pretation it is also a random variable whose expectation is given by
E
ψ
(x) =
R
3
x[ψ(x, t)[
2
d
3
x. (2.3)
In a real life setting, it will not be possible to measure x directly and one will
only be able to measure certain functions of x. For example, it is possible to
check whether the particle is inside a certain area Ω of space (e.g., inside a
detector). The corresponding observable is the characteristic function χ
Ω
(x)
of this set. In particular, the number
E
ψ
(χ
Ω
) =
R
3
χ
Ω
(x)[ψ(x, t)[
2
d
3
x =
Ω
[ψ(x, t)[
2
d
3
x (2.4)
47
48 2. Selfadjointness and spectrum
corresponds to the probability of ﬁnding the particle inside Ω ⊆ R
3
. An
important point to observe is that, in contradistinction to classical mechan
ics, the particle is no longer localized at a certain point. In particular,
the meansquare deviation (or variance) ∆
ψ
(x)
2
= E
ψ
(x
2
) − E
ψ
(x)
2
is
always nonzero.
In general, the conﬁguration space (or phase space) of a quantum
system is a (complex) Hilbert space H and the possible states of this system
are represented by the elements ψ having norm one, ψ = 1.
An observable a corresponds to a linear operator A in this Hilbert space
and its expectation, if the system is in the state ψ, is given by the real
number
E
ψ
(A) = 'ψ, Aψ` = 'Aψ, ψ`, (2.5)
where '., ..` denotes the scalar product of H. Similarly, the meansquare
deviation is given by
∆
ψ
(A)
2
= E
ψ
(A
2
) −E
ψ
(A)
2
= (A−E
ψ
(A))ψ
2
. (2.6)
Note that ∆
ψ
(A) vanishes if and only if ψ is an eigenstate corresponding to
the eigenvalue E
ψ
(A), that is, Aψ = E
ψ
(A)ψ.
From a physical point of view, (2.5) should make sense for any ψ ∈ H.
However, this is not in the cards as our simple example of one particle already
shows. In fact, the reader is invited to ﬁnd a square integrable function ψ(x)
for which xψ(x) is no longer square integrable. The deeper reason behind
this nuisance is that E
ψ
(x) can attain arbitrary large values if the particle
is not conﬁned to a ﬁnite domain, which renders the corresponding opera
tor unbounded. But unbounded operators cannot be deﬁned on the entire
Hilbert space in a natural way by the closed graph theorem (Theorem 2.7
below) .
Hence, A will only be deﬁned on a subset D(A) ⊆ H called the domain
of A. Since we want A to at least be deﬁned for most states, we require
D(A) to be dense.
However, it should be noted that there is no general prescription how to
ﬁnd the operator corresponding to a given observable.
Now let us turn to the time evolution of such a quantum mechanical
system. Given an initial state ψ(0) of the system, there should be a unique
ψ(t) representing the state of the system at time t ∈ R. We will write
ψ(t) = U(t)ψ(0). (2.7)
Moreover, it follows from physical experiments, that superposition of
states holds, that is, U(t)(α
1
ψ
1
(0) +α
2
ψ
2
(0)) = α
1
ψ
1
(t) +α
2
ψ
2
(t) ([α
1
[
2
+
[α
2
[
2
= 1). In other words, U(t) should be a linear operator. Moreover,
2.1. Some quantum mechanics 49
since ψ(t) is a state (i.e., ψ(t) = 1), we have
U(t)ψ = ψ. (2.8)
Such operators are called unitary. Next, since we have assumed uniqueness
of solutions to the initial value problem, we must have
U(0) = I, U(t +s) = U(t)U(s). (2.9)
A family of unitary operators U(t) having this property is called a one
parameter unitary group. In addition, it is natural to assume that this
group is strongly continuous
lim
t→t
0
U(t)ψ = U(t
0
)ψ, ψ ∈ H. (2.10)
Each such group has an inﬁnitesimal generator deﬁned by
Hψ = lim
t→0
i
t
(U(t)ψ −ψ), D(H) = ¦ψ ∈ H[ lim
t→0
1
t
(U(t)ψ −ψ) exists¦.
(2.11)
This operator is called the Hamiltonian and corresponds to the energy of
the system. If ψ(0) ∈ D(H), then ψ(t) is a solution of the Schr¨odinger
equation (in suitable units)
i
d
dt
ψ(t) = Hψ(t). (2.12)
This equation will be the main subject of our course.
In summary, we have the following axioms of quantum mechanics.
Axiom 1. The conﬁguration space of a quantum system is a complex
separable Hilbert space H and the possible states of this system are repre
sented by the elements of H which have norm one.
Axiom 2. Each observable a corresponds to a linear operator A deﬁned
maximally on a dense subset D(A). Moreover, the operator correspond
ing to a polynomial P
n
(a) =
¸
n
j=0
α
j
a
j
, α
j
∈ R, is P
n
(A) =
¸
n
j=0
α
j
A
j
,
D(P
n
(A)) = D(A
n
) = ¦ψ ∈ D(A)[Aψ ∈ D(A
n−1
)¦ (A
0
= I).
Axiom 3. The expectation value for a measurement of a, when the
system is in the state ψ ∈ D(A), is given by (2.5), which must be real for
all ψ ∈ D(A).
Axiom 4. The time evolution is given by a strongly continuous one
parameter unitary group U(t). The generator of this group corresponds to
the energy of the system.
In the following sections we will try to draw some mathematical conse
quences from these assumptions:
50 2. Selfadjointness and spectrum
First we will see that Axiom 2 and 3 imply that observables correspond
to selfadjoint operators. Hence these operators play a central role in quan
tum mechanics and we will derive some of their basic properties. Another
crucial role is played by the set of all possible expectation values for the
measurement of a, which is connected with the spectrum σ(A) of the corre
sponding operator A.
The problem of deﬁning functions of an observable will lead us to the
spectral theorem (in the next chapter), which generalizes the diagonalization
of symmetric matrices.
Axiom 4 will be the topic of Chapter 5.
2.2. Selfadjoint operators
Let H be a (complex separable) Hilbert space. A linear operator is a linear
mapping
A : D(A) →H, (2.13)
where D(A) is a linear subspace of H, called the domain of A. It is called
bounded if the operator norm
A = sup
ψ=1
Aψ = sup
ϕ=ψ=1
['ψ, Aϕ`[ (2.14)
is ﬁnite. The second equality follows since equality in ['ψ, Aϕ`[ ≤ ψ Aϕ
is attained when Aϕ = zψ for some z ∈ C. If A is bounded it is no restriction
to assume D(A) = H and we will always do so. The Banach space of all
bounded linear operators is denoted by L(H).
The expression 'ψ, Aψ` encountered in the previous section is called the
quadratic form
q
A
(ψ) = 'ψ, Aψ`, ψ ∈ D(A), (2.15)
associated to A. An operator can be reconstructed from its quadratic form
via the polarization identity
'ϕ, Aψ` =
1
4
(q
A
(ϕ +ψ) −q
A
(ϕ −ψ) + iq
A
(ϕ −iψ) −iq
A
(ϕ + iψ)) . (2.16)
A densely deﬁned linear operator A is called symmetric (or Hermitian)
if
'ϕ, Aψ` = 'Aϕ, ψ`, ψ, ϕ ∈ D(A). (2.17)
The justiﬁcation for this deﬁnition is provided by the following
Lemma 2.1. A densely deﬁned operator A is symmetric if and only if the
corresponding quadratic form is realvalued.
2.2. Selfadjoint operators 51
Proof. Clearly (2.17) implies that Im(q
A
(ψ)) = 0. Conversely, taking the
imaginary part of the identity
q
A
(ψ + iϕ) = q
A
(ψ) +q
A
(ϕ) + i('ψ, Aϕ` −'ϕ, Aψ`) (2.18)
shows Re'Aϕ, ψ` = Re'ϕ, Aψ`. Replacing ϕ by iϕ in this last equation
shows Im'Aϕ, ψ` = Im'ϕ, Aψ` and ﬁnishes the proof.
In other words, a densely deﬁned operator A is symmetric if and only if
'ψ, Aψ` = 'Aψ, ψ`, ψ ∈ D(A). (2.19)
This already narrows the class of admissible operators to the class of
symmetric operators by Axiom 3. Next, let us tackle the issue of the correct
domain.
By Axiom 2, A should be deﬁned maximally, that is, if
˜
A is another
symmetric operator such that A ⊆
˜
A, then A =
˜
A. Here we write A ⊆
˜
A if
D(A) ⊆ D(
˜
A) and Aψ =
˜
Aψ for all ψ ∈ D(A). In addition, we write A =
˜
A
if both
˜
A ⊆ A and A ⊆
˜
A hold.
The adjoint operator A
∗
of a densely deﬁned linear operator A is
deﬁned by
D(A
∗
) = ¦ψ ∈ H[∃
˜
ψ ∈ H : 'ψ, Aϕ` = '
˜
ψ, ϕ`, ∀ϕ ∈ D(A)¦
A
∗
ψ =
˜
ψ
. (2.20)
The requirement that D(A) is dense implies that A
∗
is welldeﬁned. How
ever, note that D(A
∗
) might not be dense in general. In fact, it might
contain no vectors other than 0.
Clearly we have (αA)
∗
= α
∗
A
∗
for α ∈ C and (A + B)
∗
⊇ A
∗
+ B
∗
provided D(A + B) = D(A) ∩ D(B) is dense. However, equality will not
hold in general unless one operator is bounded (Problem 2.1).
For later use, note that (Problem 2.2)
Ker(A
∗
) = Ran(A)
⊥
. (2.21)
For symmetric operators we clearly have A ⊆ A
∗
. If in addition, A = A
∗
holds, then A is called selfadjoint. Our goal is to show that observables
correspond to selfadjoint operators. This is for example true in the case of
the position operator x, which is a special case of a multiplication operator.
Example. (Multiplication operator) Consider the multiplication operator
(Af)(x) = A(x)f(x), D(A) = ¦f ∈ L
2
(R
n
, dµ) [ Af ∈ L
2
(R
n
, dµ)¦,
(2.22)
given by multiplication with the measurable function A : R
n
→ C. First
of all note that D(A) is dense. In fact, consider Ω
n
= ¦x ∈ R
n
[ [A(x)[ ≤
n¦ R
n
. Then, for every f ∈ L
2
(R
n
, dµ) the function f
n
= χ
Ωn
f ∈ D(A)
converges to f as n → ∞ by dominated convergence.
52 2. Selfadjointness and spectrum
Next, let us compute the adjoint of A. Performing a formal computation
we have for h, f ∈ D(A) that
'h, Af` =
h(x)
∗
A(x)f(x)dµ(x) =
(A(x)
∗
h(x))
∗
f(x)dµ(x) = '
˜
Ah, f`,
(2.23)
where
˜
A is multiplication by A(x)
∗
,
(
˜
Af)(x) = A(x)
∗
f(x), D(
˜
A) = ¦f ∈ L
2
(R
n
, dµ) [
˜
Af ∈ L
2
(R
n
, dµ)¦.
(2.24)
Note D(
˜
A) = D(A). At ﬁrst sight this seems to show that the adjoint of
A is
˜
A. But for our calculation we had to assume h ∈ D(A) and there
might be some functions in D(A
∗
) which do not satisfy this requirement! In
particular, our calculation only shows
˜
A ⊆ A
∗
. To show that equality holds,
we need to work a little harder:
If h ∈ D(A
∗
) there is some g ∈ L
2
(R
n
, dµ) such that
h(x)
∗
A(x)f(x)dµ(x) =
g(x)
∗
f(x)dµ(x), f ∈ D(A), (2.25)
and thus
(h(x)A(x)
∗
−g(x))
∗
f(x)dµ(x) = 0, f ∈ D(A). (2.26)
In particular,
χ
Ωn
(x)(h(x)A(x)
∗
−g(x))
∗
f(x)dµ(x) = 0, f ∈ L
2
(R
n
, dµ), (2.27)
which shows that χ
Ωn
(h(x)A(x)
∗
− g(x))
∗
∈ L
2
(R
n
, dµ) vanishes. Since n
is arbitrary, we even have h(x)A(x)
∗
= g(x) ∈ L
2
(R
n
, dµ) and thus A
∗
is
multiplication by A(x)
∗
and D(A
∗
) = D(A).
In particular, A is selfadjoint if A is realvalued. In the general case we
have at least Af = A
∗
f for all f ∈ D(A) = D(A
∗
). Such operators are
called normal.
Now note that
A ⊆ B ⇒ B
∗
⊆ A
∗
, (2.28)
that is, increasing the domain of A implies decreasing the domain of A
∗
.
Thus there is no point in trying to extend the domain of a selfadjoint
operator further. In fact, if A is selfadjoint and B is a symmetric extension,
we infer A ⊆ B ⊆ B
∗
⊆ A
∗
= A implying A = B.
Corollary 2.2. Selfadjoint operators are maximal, that is, they do not have
any symmetric extensions.
Furthermore, if A
∗
is densely deﬁned (which is the case if A is symmetric)
we can consider A
∗∗
. From the deﬁnition (2.20) it is clear that A ⊆ A
∗∗
and
2.2. Selfadjoint operators 53
thus A
∗∗
is an extension of A. This extension is closely related to extending
a linear subspace M via M
⊥⊥
= M (as we will see a bit later) and thus is
called the closure A = A
∗∗
of A.
If A is symmetric we have A ⊆ A
∗
and hence A = A
∗∗
⊆ A
∗
, that is,
A lies between A and A
∗
. Moreover, 'ψ, A
∗
ϕ` = 'Aψ, ϕ` for all ψ ∈ D(A),
ϕ ∈ D(A
∗
) implies that A is symmetric since A
∗
ϕ = Aϕ for ϕ ∈ D(A).
Example. (Diﬀerential operator) Take H = L
2
(0, 2π).
(i). Consider the operator
A
0
f = −i
d
dx
f, D(A
0
) = ¦f ∈ C
1
[0, 2π] [ f(0) = f(2π) = 0¦. (2.29)
That A
0
is symmetric can be shown by a simple integration by parts (do
this). Note that the boundary conditions f(0) = f(2π) = 0 are chosen
such that the boundary terms occurring from integration by parts vanish.
However, this will also follow once we have computed A
∗
0
. If g ∈ D(A
∗
0
) we
must have
2π
0
g(x)
∗
(−if
(x))dx =
2π
0
˜ g(x)
∗
f(x)dx (2.30)
for some ˜ g ∈ L
2
(0, 2π). Integration by parts shows
2π
0
f
(x)
g(x) −i
x
0
˜ g(t)dt
∗
dx = 0 (2.31)
and hence g(x) − i
x
0
˜ g(t)dt ∈ ¦f
[f ∈ D(A
0
)¦
⊥
. But ¦f
[f ∈ D(A
0
)¦ =
¦h ∈ C(0, 2π)[
2π
0
h(t)dt = 0¦ implying g(x) = g(0) + i
x
0
˜ g(t)dt since
¦f
[f ∈ D(A
0
)¦ = ¦h ∈ H['1, h` = 0¦ = ¦1¦
⊥
and ¦1¦
⊥⊥
= span¦1¦. Thus
g ∈ AC[0, 2π], where
AC[a, b] = ¦f ∈ C[a, b][f(x) = f(a) +
x
a
g(t)dt, g ∈ L
1
(a, b)¦ (2.32)
denotes the set of all absolutely continuous functions (see Section 2.6). In
summary, g ∈ D(A
∗
0
) implies g ∈ AC[0, 2π] and A
∗
0
g = ˜ g = −ig
. Conversely,
for every g ∈ H
1
(0, 2π) = ¦f ∈ AC[0, 2π][f
∈ L
2
(0, 2π)¦ (2.30) holds with
˜ g = −ig
and we conclude
A
∗
0
f = −i
d
dx
f, D(A
∗
0
) = H
1
(0, 2π). (2.33)
In particular, A is symmetric but not selfadjoint. Since A
∗∗
⊆ A
∗
we
compute
0 = 'g, A
0
f` −'A
∗
0
g, f` = i(f(0)g(0)
∗
−f(2π)g(2π)
∗
) (2.34)
54 2. Selfadjointness and spectrum
and since the boundary values of g ∈ D(A
∗
0
) can be prescribed arbitrary, we
must have f(0) = f(2π) = 0. Thus
A
0
f = −i
d
dx
f, D(A
0
) = ¦f ∈ D(A
∗
0
) [ f(0) = f(2π) = 0¦. (2.35)
(ii). Now let us take
Af = −i
d
dx
f, D(A) = ¦f ∈ C
1
[0, 2π] [ f(0) = f(2π)¦. (2.36)
which is clearly an extension of A
0
. Thus A
∗
⊆ A
∗
0
and we compute
0 = 'g, Af` −'A
∗
g, f` = if(0)(g(0)
∗
−g(2π)
∗
). (2.37)
Since this must hold for all f ∈ D(A) we conclude g(0) = g(2π) and
A
∗
f = −i
d
dx
f, D(A
∗
) = ¦f ∈ H
1
(0, 2π) [ f(0) = f(2π)¦. (2.38)
Similarly, as before, A = A
∗
and thus A is selfadjoint.
One might suspect that there is no big diﬀerence between the two sym
metric operators A
0
and A from the previous example, since they coincide
on a dense set of vectors. However, the converse is true: For example, the
ﬁrst operator A
0
has no eigenvectors at all (i.e., solutions of the equation
A
0
ψ = zψ, z ∈ C) whereas the second one has an orthonormal basis of
eigenvectors!
Example. Compute the eigenvectors of A
0
and A from the previous exam
ple.
(i). By deﬁnition an eigenvector is a (nonzero) solution of A
0
u = zu,
z ∈ C, that is, a solution of the ordinary diﬀerential equation
u
(x) = zu(x) (2.39)
satisfying the boundary conditions u(0) = u(2π) = 0 (since we must have
u ∈ D(A
0
). The general solution of the diﬀerential equation is u(x) =
u(0)e
izx
and the boundary conditions imply u(x) = 0. Hence there are no
eigenvectors.
(ii). Now we look for solutions of Au = zu, that is the same diﬀerential
equation as before, but now subject to the boundary condition u(0) = u(2π).
Again the general solution is u(x) = u(0)e
izx
and the boundary condition
requires u(0) = u(0)e
2πiz
. Thus there are two possibilities. Either u(0) = 0
(which is of no use for us) or z ∈ Z. In particular, we see that all eigenvectors
are given by
u
n
(x) =
1
√
2π
e
inx
, n ∈ Z, (2.40)
which are wellknown to form an orthonormal basis.
2.2. Selfadjoint operators 55
We will see a bit later that this is a consequence of selfadjointness of
A. Hence it will be important to know whether a given operator is self
adjoint or not. Our example shows that symmetry is easy to check (in case
of diﬀerential operators it usually boils down to integration by parts), but
computing the adjoint of an operator is a nontrivial job even in simple situ
ations. However, we will learn soon that selfadjointness is a much stronger
property than symmetry justifying the additional eﬀort needed to prove it.
On the other hand, if a given symmetric operator A turns out not to
be selfadjoint, this raises the question of selfadjoint extensions. Two cases
need to be distinguished. If A is selfadjoint, then there is only one self
adjoint extension (if B is another one, we have A ⊆ B and hence A = B
by Corollary 2.2). In this case A is called essentially selfadjoint and
D(A) is called a core for A. Otherwise there might be more than one self
adjoint extension or none at all. This situation is more delicate and will be
investigated in Section 2.5.
Since we have seen that computing A
∗
is not always easy, a criterion for
selfadjointness not involving A
∗
will be useful.
Lemma 2.3. Let A be symmetric such that Ran(A+z) = Ran(A+z
∗
) = H
for one z ∈ C. Then A is selfadjoint.
Proof. Let ψ ∈ D(A
∗
) and A
∗
ψ =
˜
ψ. Since Ran(A + z
∗
) = H, there is a
ϑ ∈ D(A) such that (A+z
∗
)ϑ =
˜
ψ +z
∗
ψ. Now we compute
'ψ, (A+z)ϕ` = '
˜
ψ +z
∗
ψ, ϕ` = '(A+z
∗
)ϑ, ϕ` = 'ϑ, (A+z)ϕ`, ϕ ∈ D(A),
(2.41)
and hence ψ = ϑ ∈ D(A) since Ran(A+z) = H.
To proceed further, we will need more information on the closure of
an operator. We will use a diﬀerent approach which avoids the use of the
adjoint operator. We will establish equivalence with our original deﬁnition
in Lemma 2.4.
The simplest way of extending an operator A is to take the closure of its
graph Γ(A) = ¦(ψ, Aψ)[ψ ∈ D(A)¦ ⊂ H
2
. That is, if (ψ
n
, Aψ
n
) → (ψ,
˜
ψ)
we might try to deﬁne Aψ =
˜
ψ. For Aψ to be welldeﬁned, we need that
(ψ
n
, Aψ
n
) → (0,
˜
ψ) implies
˜
ψ = 0. In this case A is called closable and the
unique operator A which satisﬁes Γ(A) = Γ(A) is called the closure of A.
Clearly, A is called closed if A = A, which is the case if and only if the
graph of A is closed. Equivalently, A is closed if and only if Γ(A) equipped
with the graph norm ψ
2
Γ(A)
= ψ
2
+ Aψ
2
is a Hilbert space (i.e.,
closed). A bounded operator is closed if and only if its domain is closed
(show this!).
56 2. Selfadjointness and spectrum
Example. Let us compute the closure of the operator A
0
from the pre
vious example without the use of the adjoint operator. Let f ∈ D(A
0
)
and let f
n
∈ D(A
0
) be a sequence such that f
n
→ f, A
0
f
n
→ −ig. Then
f
n
→ g and hence f(x) =
x
0
g(t)dt. Thus f ∈ AC[0, 2π] and f(0) = 0.
Moreover f(2π) = lim
n→0
2π
0
f
n
(t)dt = 0. Conversely, any such f can be
approximated by functions in D(A
0
) (show this).
Next, let us collect a few important results.
Lemma 2.4. Suppose A is a densely deﬁned operator.
(i) A
∗
is closed.
(ii) A is closable if and only if D(A
∗
) is dense and A = A
∗∗
respectively
(A)
∗
= A
∗
in this case.
(iii) If A is injective and the Ran(A) is dense, then (A
∗
)
−1
= (A
−1
)
∗
.
If A is closable and A is injective, then A
−1
= A
−1
.
Proof. Let us consider the following two unitary operators from H
2
to itself
U(ϕ, ψ) = (ψ, −ϕ), V (ϕ, ψ) = (ψ, ϕ). (2.42)
(i). From
Γ(A
∗
) = ¦(ϕ, ˜ ϕ) ∈ H
2
['ϕ, Aψ` = ' ˜ ϕ, ψ` ∀ϕ ∈ D(A
∗
)¦
= ¦(ϕ, ˜ ϕ) ∈ H
2
['(−˜ ϕ, ϕ), (ψ,
˜
ψ)`
Γ(A)
= 0 ∀(ψ,
˜
ψ) ∈ Γ(A)¦
= U(Γ(A)
⊥
) = (UΓ(A))
⊥
(2.43)
we conclude that A
∗
is closed.
(ii). From
Γ(A) = Γ(A)
⊥⊥
= (UΓ(A
∗
))
⊥
= ¦(ψ,
˜
ψ)[ 'ψ, A
∗
ϕ` −'
˜
ψ, ϕ` = 0, ∀ϕ ∈ D(A
∗
)¦ (2.44)
we see that (0,
˜
ψ) ∈ Γ(A) if and only if
˜
ψ ∈ D(A
∗
)
⊥
. Hence A is closable if
and only if D(A
∗
) is dense. In this case, equation (2.43) also shows A
∗
= A
∗
.
Moreover, replacing A by A
∗
in (2.43) and comparing with the last formula
shows A
∗∗
= A.
(iii). Next note that (provided A is injective)
Γ(A
−1
) = V Γ(A). (2.45)
Hence if Ran(A) is dense, then Ker(A
∗
) = Ran(A)
⊥
= ¦0¦ and
Γ((A
∗
)
−1
) = V Γ(A
∗
) = V UΓ(A)
⊥
= UV Γ(A)
⊥
= U(V Γ(A))
⊥
(2.46)
2.2. Selfadjoint operators 57
shows that (A
∗
)
−1
= (A
−1
)
∗
. Similarly, if A is closable and A is injective,
then A
−1
= A
−1
by
Γ(A
−1
) = V Γ(A) = V Γ(A) = Γ(A
−1
). (2.47)
If A ∈ L(H) we clearly have D(A
∗
) = H and by Corollary 1.8 A ∈ L(H).
In particular, since A = A
∗∗
we obtain
Theorem 2.5. We have A ∈ L(H) if and only if A
∗
∈ L(H).
Now we can also generalize Lemma 2.3 to the case of essential selfadjoint
operators.
Lemma 2.6. A symmetric operator A is essentially selfadjoint if and only
if one of the following conditions holds for one z ∈ C`R.
• Ran(A+z) = Ran(A+z
∗
) = H.
• Ker(A
∗
+z) = Ker(A
∗
+z
∗
) = ¦0¦.
If A is nonnegative, that is 'ψ, Aψ` ≥ 0 for all ψ ∈ D(A), we can also
admit z ∈ (−∞, 0).
Proof. As noted earlier Ker(A
∗
) = Ran(A)
⊥
, and hence the two conditions
are equivalent. By taking the closure of A it is no restriction to assume that
A is closed. Let z = x + iy. From
(A−z)ψ
2
= (A−x)ψ−iyψ
2
= (A−x)ψ
2
+y
2
ψ
2
≥ y
2
ψ
2
, (2.48)
we infer that Ker(A−z) = ¦0¦ and hence (A−z)
−1
exists. Moreover, setting
ψ = (A − z)
−1
ϕ (y = 0) shows (A − z)
−1
 ≤ [y[
−1
. Hence (A − z)
−1
is
bounded and closed. Since it is densely deﬁned by assumption, its domain
Ran(A+z) must be equal to H. Replacing z by z
∗
and applying Lemma 2.3
ﬁnishes the general case. The argument for the nonnegative case with
z < 0 is similar using εψ
2
≤ ['ψ, (A + ε)ψ`[
2
≤ ψ(A + ε)ψ which
shows (A+ε)
−1
≤ ε
−1
, ε > 0.
In addition, we can also prove the closed graph theorem which shows
that an unbounded operator cannot be deﬁned on the entire Hilbert space.
Theorem 2.7 (closed graph). Let H
1
and H
2
be two Hilbert spaces and A
an operator deﬁned on all of H
1
. Then A is bounded if and only if Γ(A) is
closed.
Proof. If A is bounded than it is easy to see that Γ(A) is closed. So let us
assume that Γ(A) is closed. Then A
∗
is well deﬁned and for all unit vectors
58 2. Selfadjointness and spectrum
ϕ ∈ D(A
∗
) we have that the linear functional
ϕ
(ψ) = 'A
∗
ϕ, ψ` is pointwise
bounded

ϕ
(ψ) = ['ϕ, Aψ`[ ≤ Aψ. (2.49)
Hence by the uniform boundedness principle there is a constant C such that

ϕ
 = A
∗
ϕ ≤ C. That is, A
∗
is bounded and so is A = A
∗∗
.
Finally we want to draw some some further consequences of Axiom 2
and show that observables correspond to selfadjoint operators. Since self
adjoint operators are already maximal, the diﬃcult part remaining is to
show that an observable has at least one selfadjoint extension. There is a
good way of doing this for nonnegative operators and hence we will consider
this case ﬁrst.
An operator is called nonnegative (resp. positive) if 'ψ, Aψ` ≥ 0
(resp. > 0 for ψ = 0) for all ψ ∈ D(A). If A is positive, the map (ϕ, ψ) →
'ϕ, Aψ` is a scalar product. However, there might be sequences which are
Cauchy with respect to this scalar product but not with respect to our
original one. To avoid this, we introduce the scalar product
'ϕ, ψ`
A
= 'ϕ, (A+ 1)ψ`, A ≥ 0, (2.50)
deﬁned on D(A), which satisﬁes ψ ≤ ψ
A
. Let H
A
be the completion of
D(A) with respect to the above scalar product. We claim that H
A
can be
regarded as a subspace of H, that is, D(A) ⊆ H
A
⊆ H.
If (ψ
n
) is a Cauchy sequence in D(A), then it is also Cauchy in H (since
ψ ≤ ψ
A
by assumption) and hence we can identify it with the limit of
(ψ
n
) regarded as a sequence in H. For this identiﬁcation to be unique, we
need to show that if (ψ
n
) ⊂ D(A) is a Cauchy sequence in H
A
such that
ψ
n
 → 0, then ψ
n

A
→ 0. This follows from
ψ
n

2
A
= 'ψ
n
, ψ
n
−ψ
m
`
A
+'ψ
n
, ψ
m
`
A
≤ ψ
n

A
ψ
n
−ψ
m

A
+ψ
n
(A+ 1)ψ
m
 (2.51)
since the right hand side can be made arbitrarily small choosing m, n large.
Clearly the quadratic form q
A
can be extended to every ψ ∈ H
A
by
setting
q
A
(ψ) = 'ψ, ψ`
A
−ψ
2
, ψ ∈ Q(A) = H
A
. (2.52)
The set Q(A) is also called the form domain of A.
Example. (Multiplication operator) Let A be multiplication by A(x) ≥ 0
in L
2
(R
n
, dµ). Then
Q(A) = D(A
1/2
) = ¦f ∈ L
2
(R
n
, dµ) [ A
1/2
f ∈ L
2
(R
n
, dµ)¦ (2.53)
and
q
A
(x) =
R
n
A(x)[f(x)[
2
dµ(x) (2.54)
2.2. Selfadjoint operators 59
(show this).
Now we come to our extension result. Note that A + 1 is injective and
the best we can hope for is that for a nonnegative extension
˜
A,
˜
A + 1 is a
bijection from D(
˜
A) onto H.
Lemma 2.8. Suppose A is a nonnegative operator, then there is a non
negative extension
˜
A such that Ran(
˜
A+ 1) = H.
Proof. Let us deﬁne an operator
˜
A by
D(
˜
A) = ¦ψ ∈ H
A
[∃
˜
ψ ∈ H : 'ϕ, ψ`
A
= 'ϕ,
˜
ψ`, ∀ϕ ∈ H
A
¦
˜
Aψ =
˜
ψ −ψ
. (2.55)
Since H
A
is dense,
˜
ψ is welldeﬁned. Moreover, it is straightforward to see
that
˜
A is a nonnegative extension of A.
It is also not hard to see that Ran(
˜
A + 1) = H. Indeed, for any
˜
ψ ∈ H,
ϕ → '
˜
ψ, ϕ` is bounded linear functional on H
A
. Hence there is an element
ψ ∈ H
A
such that '
˜
ψ, ϕ` = 'ψ, ϕ`
A
for all ϕ ∈ H
A
. By the deﬁnition of
˜
A,
(
˜
A+ 1)ψ =
˜
ψ and hence
˜
A+ 1 is onto.
Now it is time for another
Example. Let us take H = L
2
(0, π) and consider the operator
Af = −
d
2
dx
2
f, D(A) = ¦f ∈ C
2
[0, π] [ f(0) = f(π) = 0¦, (2.56)
which corresponds to the onedimensional model of a particle conﬁned to a
box.
(i). First of all, using integration by parts twice, it is straightforward to
check that A is symmetric
π
0
g(x)
∗
(−f
)(x)dx =
π
0
g
(x)
∗
f
(x)dx =
π
0
(−g
)(x)
∗
f(x)dx. (2.57)
Note that the boundary conditions f(0) = f(π) = 0 are chosen such that
the boundary terms occurring from integration by parts vanish. Moreover,
the same calculation also shows that A is positive
π
0
f(x)
∗
(−f
)(x)dx =
π
0
[f
(x)[
2
dx > 0, f = 0. (2.58)
(ii). Next let us show H
A
= ¦f ∈ H
1
(0, π) [ f(0) = f(π) = 0¦. In fact,
since
'g, f`
A
=
π
0
g
(x)
∗
f
(x) +g(x)
∗
f(x)
dx, (2.59)
we see that f
n
is Cauchy in H
A
if and only if both f
n
and f
n
are Cauchy
in L
2
(0, π). Thus f
n
→ f and f
n
→ g in L
2
(0, π) and f
n
(x) =
x
0
f
n
(t)dt
60 2. Selfadjointness and spectrum
implies f(x) =
x
0
g(t)dt. Thus f ∈ AC[0, π]. Moreover, f(0) = 0 is obvious
and from 0 = f
n
(π) =
π
0
f
n
(t)dt we have f(π) = lim
n→∞
π
0
f
n
(t)dt = 0.
So we have H
A
⊆ ¦f ∈ H
1
(0, π) [ f(0) = f(π) = 0¦. To see the converse
approximate f
by smooth functions g
n
. Using g
n
−
π
0
g
n
(t)dt instead of g
n
it is no restriction to assume
π
0
g
n
(t)dt = 0. Now deﬁne f
n
(x) =
x
0
g
n
(t)dt
and note f
n
∈ D(A) → f.
(iii). Finally, let us compute the extension
˜
A. We have f ∈ D(
˜
A) if for
all g ∈ H
A
there is an
˜
f such that 'g, f`
A
= 'g,
˜
f`. That is,
π
0
g
(x)
∗
f
(x)dx =
π
0
g(x)
∗
(
˜
f(x) −f(x))dx. (2.60)
Integration by parts on the right hand side shows
π
0
g
(x)
∗
f
(x)dx = −
π
0
g
(x)
∗
x
0
(
˜
f(t) −f(t))dt dx (2.61)
or equivalently
π
0
g
(x)
∗
f
(x) +
x
0
(
˜
f(t) −f(t))dt
dx = 0. (2.62)
Now observe ¦g
∈ H[g ∈ H
A
¦ = ¦h ∈ H[
π
0
h(t)dt = 0¦ = ¦1¦
⊥
and thus
f
(x) +
x
0
(
˜
f(t) − f(t))dt ∈ ¦1¦
⊥⊥
= span¦1¦. So we see f ∈ H
2
(0, π) =
¦f ∈ AC[0, π][f
∈ H
1
(0, π)¦ and
˜
Af = −f
. The converse is easy and
hence
˜
Af = −
d
2
dx
2
f, D(
˜
A) = ¦f ∈ H
2
[0, π] [ f(0) = f(π) = 0¦. (2.63)
Now let us apply this result to operators A corresponding to observ
ables. Since A will, in general, not satisfy the assumptions of our lemma, we
will consider 1 + A
2
, D(1 + A
2
) = D(A
2
), instead, which has a symmetric
extension whose range is H. By our requirement for observables, 1 + A
2
is maximally deﬁned and hence is equal to this extension. In other words,
Ran(1 +A
2
) = H. Moreover, for any ϕ ∈ H there is a ψ ∈ D(A
2
) such that
(A−i)(A+ i)ψ = (A+ i)(A−i)ψ = ϕ (2.64)
and since (A ± i)ψ ∈ D(A), we infer Ran(A ± i) = H. As an immediate
consequence we obtain
Corollary 2.9. Observables correspond to selfadjoint operators.
But there is another important consequence of the results which is worth
while mentioning.
2.3. Resolvents and spectra 61
Theorem 2.10 (Friedrichs extension). Let A be a semibounded symmet
ric operator, that is,
q
A
(ψ) = 'ψ, Aψ` ≥ γψ
2
, γ ∈ R. (2.65)
Then there is a selfadjoint extension
˜
A which is also bounded from below
by γ and which satisﬁes D(
˜
A) ⊆ H
A−γ
.
Proof. Replacing A by A − γ we can reduce it to the case considered in
Lemma 2.8. The rest is straightforward.
Problem 2.1. Show (αA)
∗
= α
∗
A
∗
and (A+B)
∗
⊇ A
∗
+B
∗
(where D(A
∗
+
B
∗
) = D(A
∗
) ∩ D(B
∗
)) with equality if one operator is bounded. Give an
example where equality does not hold.
Problem 2.2. Show (2.21).
Problem 2.3. Show that if A is normal, so is A+z for any z ∈ C.
Problem 2.4. Show that normal operators are closed.
Problem 2.5. Show that the kernel of a closed operator is closed.
Problem 2.6. Show that if A is bounded and B closed, then BA is closed.
Problem 2.7. Let A = −
d
2
dx
2
, D(A) = ¦f ∈ H
2
(0, π) [ f(0) = f(π) = 0¦
and let ψ(x) =
1
2
√
π
x(π−x). Find the error in the following argument: Since
A is symmetric we have 1 = 'Aψ, Aψ` = 'ψ, A
2
ψ` = 0.
Problem 2.8. Suppose A is a closed operator. Show that A
∗
A (with D(A
∗
A) =
¦ψ ∈ D(A)[Aψ ∈ D(A
∗
)¦ is selfadjoint. (Hint: A
∗
A ≥ 0.)
Problem 2.9. Show that A is normal if and only if AA
∗
= A
∗
A.
2.3. Resolvents and spectra
Let A be a (densely deﬁned) closed operator. The resolvent set of A is
deﬁned by
ρ(A) = ¦z ∈ C[(A−z)
−1
∈ L(H)¦. (2.66)
More precisely, z ∈ ρ(A) if and only if (A − z) : D(A) → H is bijective
and its inverse is bounded. By the closed graph theorem (Theorem 2.7), it
suﬃces to check that A − z is bijective. The complement of the resolvent
set is called the spectrum
σ(A) = C`ρ(A) (2.67)
of A. In particular, z ∈ σ(A) if A − z has a nontrivial kernel. A vector
ψ ∈ Ker(A−z) is called an eigenvector and z is called eigenvalue in this
case.
62 2. Selfadjointness and spectrum
The function
R
A
: ρ(A) → L(H)
z → (A−z)
−1
(2.68)
is called resolvent of A. Note the convenient formula
R
A
(z)
∗
= ((A−z)
−1
)
∗
= ((A−z)
∗
)
−1
= (A
∗
−z
∗
)
−1
= R
A
∗(z
∗
). (2.69)
In particular,
ρ(A
∗
) = ρ(A)
∗
. (2.70)
Example. (Multiplication operator) Consider again the multiplication op
erator
(Af)(x) = A(x)f(x), D(A) = ¦f ∈ L
2
(R
n
, dµ) [ Af ∈ L
2
(R
n
, dµ)¦,
(2.71)
given by multiplication with the measurable function A : R
n
→ C. Clearly
(A−z)
−1
is given by the multiplication operator
(A−z)
−1
f(x) =
1
A(x) −z
f(x),
D((A−z)
−1
) = ¦f ∈ L
2
(R
n
, dµ) [
1
A−z
f ∈ L
2
(R
n
, dµ)¦ (2.72)
whenever this operator is bounded. But (A − z)
−1
 = 
1
A−z

∞
≤
1
ε
is
equivalent to µ(¦x[ [A(x) −z[ ≥ ε¦) = 0 and hence
ρ(A) = ¦z ∈ C[∃ε > 0 : µ(¦x[ [A(x) −z[ ≤ ε¦) = 0¦. (2.73)
Moreover, z is an eigenvalue of A if µ(A
−1
(¦z¦)) > 0 and χ
A
−1
({z})
is a
corresponding eigenfunction in this case.
Example. (Diﬀerential operator) Consider again the diﬀerential operator
Af = −i
d
dx
f, D(A) = ¦f ∈ AC[0, 2π] [ f
∈ L
2
, f(0) = f(2π)¦ (2.74)
in L
2
(0, 2π). We already know that the eigenvalues of A are the integers
and that the corresponding normalized eigenfunctions
u
n
(x) =
1
√
2π
e
inx
(2.75)
form an orthonormal basis.
To compute the resolvent we must ﬁnd the solution of the correspond
ing inhomogeneous equation −if
(x) − z f(x) = g(x). By the variation of
constants formula the solution is given by (this can also be easily veriﬁed
directly)
f(x) = f(0)e
izx
+ i
x
0
e
iz(x−t)
g(t)dt. (2.76)
2.3. Resolvents and spectra 63
Since f must lie in the domain of A, we must have f(0) = f(2π) which gives
f(0) =
i
e
−2πiz
−1
2π
0
e
−izt
g(t)dt, z ∈ C`Z. (2.77)
(Since z ∈ Z are the eigenvalues, the inverse cannot exist in this case.) Hence
(A−z)
−1
g(x) =
2π
0
G(z, x, t)g(t)dt, (2.78)
where
G(z, x, t) = e
iz(x−t)
−i
1−e
−2πiz
, t > x
i
1−e
2πiz
, t < x
, z ∈ C`Z. (2.79)
In particular σ(A) = Z.
If z, z
∈ ρ(A), we have the ﬁrst resolvent formula
R
A
(z) −R
A
(z
) = (z −z
)R
A
(z)R
A
(z
) = (z −z
)R
A
(z
)R
A
(z). (2.80)
In fact,
(A−z)
−1
−(z −z
)(A−z)
−1
(A−z
)
−1
=
(A−z)
−1
(1 −(z −A+A−z
)(A−z
)
−1
) = (A−z
)
−1
, (2.81)
which proves the ﬁrst equality. The second follows after interchanging z and
z
. Now ﬁx z
= z
0
and use (2.80) recursively to obtain
R
A
(z) =
n
¸
j=0
(z −z
0
)
j
R
A
(z
0
)
j+1
+ (z −z
0
)
n+1
R
A
(z
0
)
n+1
R
A
(z). (2.82)
The sequence of bounded operators
R
n
=
n
¸
j=0
(z −z
0
)
j
R
A
(z
0
)
j+1
(2.83)
converges to a bounded operator if [z − z
0
[ < R
A
(z
0
)
−1
and clearly we
expect z ∈ ρ(A) and R
n
→ R
A
(z) in this case. Let R
∞
= lim
n→∞
R
n
and
set ϕ
n
= R
n
ψ, ϕ = R
∞
ψ for some ψ ∈ H. Then a quick calculation shows
AR
n
ψ = ψ + (z −z
0
)ϕ
n−1
+z
0
ϕ
n
. (2.84)
Hence (ϕ
n
, Aϕ
n
) → (ϕ, ψ + zϕ) shows ϕ ∈ D(A) (since A is closed) and
(A−z)R
∞
ψ = ψ. Similarly, for ψ ∈ D(A),
R
n
Aψ = ψ + (z −z
0
)ϕ
n−1
+z
0
ϕ
n
(2.85)
and hence R
∞
(A − z)ψ = ψ after taking the limit. Thus R
∞
= R
A
(z) as
anticipated.
64 2. Selfadjointness and spectrum
If A is bounded, a similar argument veriﬁes the Neumann series for
the resolvent
R
A
(z) = −
n−1
¸
j=0
A
j
z
j+1
+
1
z
n
A
n
R
A
(z)
= −
∞
¸
j=0
A
j
z
j+1
, [z[ > A. (2.86)
In summary we have proved the following
Theorem 2.11. The resolvent set ρ(A) is open and R
A
: ρ(A) → L(H) is
holomorphic, that is, it has an absolutely convergent power series expansion
around every point z
0
∈ ρ(A). In addition,
R
A
(z) ≥ dist(z, σ(A))
−1
(2.87)
and if A is bounded we have ¦z ∈ C[ [z[ > A¦ ⊆ ρ(A).
As a consequence we obtain the useful
Lemma 2.12. We have z ∈ σ(A) if there is a sequence ψ
n
∈ D(A) such
that ψ
n
 = 1 and (A−z)ψ
n
 → 0. If z is a boundary point of ρ(A), then
the converse is also true. Such a sequence is called Weyl sequence.
Proof. Let ψ
n
be a Weyl sequence. Then z ∈ ρ(A) is impossible by 1 =
ψ
n
 = R
A
(z)(A − z)ψ
n
 ≤ R
A
(z)(A − z)ψ
n
 → 0. Conversely, by
(2.87) there is a sequence z
n
→ z and corresponding vectors ϕ
n
∈ H such
that R
A
(z)ϕ
n
ϕ
n

−1
→ ∞. Let ψ
n
= R
A
(z
n
)ϕ
n
and rescale ϕ
n
such that
ψ
n
 = 1. Then ϕ
n
 → 0 and hence
(A−z)ψ
n
 = ϕ
n
+ (z
n
−z)ψ
n
 ≤ ϕ
n
 +[z −z
n
[ → 0 (2.88)
shows that ψ
n
is a Weyl sequence.
Let us also note the following spectral mapping result.
Lemma 2.13. Suppose A is injective, then
σ(A
−1
)`¦0¦ = (σ(A)`¦0¦)
−1
. (2.89)
In addition, we have Aψ = zψ if and only if A
−1
ψ = z
−1
ψ.
Proof. Suppose z ∈ ρ(A)`¦0¦. Then we claim
R
A
−1(z
−1
) = −zAR
A
(z) = −z −R
A
(z). (2.90)
In fact, the right hand side is a bounded operator from H → Ran(A) =
D(A
−1
) and
(A
−1
−z
−1
)(−zAR
A
(z))ϕ = (−z +A)R
A
(z)ϕ = ϕ, ϕ ∈ H. (2.91)
2.3. Resolvents and spectra 65
Conversely, if ψ ∈ D(A
−1
) = Ran(A) we have ψ = Aϕ and hence
(−zAR
A
(z))(A
−1
−z
−1
)ψ = AR
A
(z)((A−z)ϕ) = Aϕ = ψ. (2.92)
Thus z
−1
∈ ρ(A
−1
). The rest follows after interchanging the roles of A and
A
−1
.
Next, let us characterize the spectra of selfadjoint operators.
Theorem 2.14. Let A be symmetric. Then A is selfadjoint if and only if
σ(A) ⊆ R and A ≥ 0 if and only if σ(A) ⊆ [0, ∞). Moreover, R
A
(z) ≤
[Im(z)[
−1
and, if A ≥ 0, R
A
(λ) ≤ [λ[
−1
, λ < 0.
Proof. If σ(A) ⊆ R, then Ran(A + z) = H, z ∈ C`R, and hence A is self
adjoint by Lemma 2.6. Conversely, if A is selfadjoint (resp. A ≥ 0), then
R
A
(z) exists for z ∈ C`R (resp. z ∈ C`(−∞, 0]) and satisﬁes the given
estimates as has been shown in the proof of Lemma 2.6.
In particular, we obtain
Theorem 2.15. Let A be selfadjoint, then
inf σ(A) = inf
ψ∈D(A), ψ=1
'ψ, Aψ`. (2.93)
For the eigenvalues and corresponding eigenfunctions we have:
Lemma 2.16. Let A be symmetric. Then all eigenvalues are real and eigen
vectors corresponding to diﬀerent eigenvalues are orthogonal.
Proof. If Aψ
j
= λ
j
ψ
j
, j = 1, 2, we have
λ
1
ψ
1

2
= 'ψ
1
, λ
1
ψ
1
` = 'ψ
1
, Aψ
1
` = 'ψ
1
, Aψ
1
` = 'λ
1
ψ
1
, ψ
1
` = λ
∗
1
ψ
1

2
(2.94)
and
(λ
1
−λ
2
)'ψ
1
, ψ
2
` = 'Aψ
1
, ψ
2
` −'Aψ
1
, ψ
2
` = 0, (2.95)
ﬁnishing the proof.
The result does not imply that two linearly independent eigenfunctions
to the same eigenvalue are orthogonal. However, it is no restriction to
assume that they are since we can use GramSchmidt to ﬁnd an orthonormal
basis for Ker(A − λ). If H is ﬁnite dimensional, we can always ﬁnd an
orthonormal basis of eigenvectors. In the inﬁnite dimensional case this is
no longer true in general. However, if there is an orthonormal basis of
eigenvectors, then A is essentially selfadjoint.
Theorem 2.17. Suppose A is a symmetric operator which has an orthonor
mal basis of eigenfunctions ¦ϕ
j
¦, then A is essentially selfadjoint. In par
ticular, it is essentially selfadjoint on span¦ϕ
j
¦.
66 2. Selfadjointness and spectrum
Proof. Consider the set of all ﬁnite linear combinations ψ =
¸
n
j=0
c
j
ϕ
j
which is dense in H. Then φ =
¸
n
j=0
c
j
λ
j
±i
ϕ
j
∈ D(A) and (A ± i)φ = ψ
shows that Ran(A±i) is dense.
In addition, we note the following asymptotic expansion for the resolvent.
Lemma 2.18. Suppose A is selfadjoint. For every ψ ∈ H we have
lim
Im(z)→∞
AR
A
(z)ψ = 0. (2.96)
In particular, if ψ ∈ D(A
n
), then
R
A
(z)ψ = −
n
¸
j=0
A
j
ψ
z
j+1
+o(
1
z
n+1
), as Im(z) → ∞. (2.97)
Proof. It suﬃces to prove the ﬁrst claim since the second then follows as
in (2.86).
Write ψ =
˜
ψ +ϕ, where
˜
ψ ∈ D(A) and ϕ ≤ ε. Then
AR
A
(z)ψ ≤ R
A
(z)A
˜
ψ +AR
A
(z)ϕ
≤
A
˜
ψ
Im(z)
+ϕ, (2.98)
by (2.48), ﬁnishing the proof.
Similarly, we can characterize the spectra of unitary operators. Recall
that a bijection U is called unitary if 'Uψ, Uψ` = 'ψ, U
∗
Uψ` = 'ψ, ψ`. Thus
U is unitary if and only if
U
∗
= U
−1
. (2.99)
Theorem 2.19. Let U be unitary, then σ(U) ⊆ ¦z ∈ C[ [z[ = 1¦. All
eigenvalues have modulus one and eigenvectors corresponding to diﬀerent
eigenvalues are orthogonal.
Proof. Since U ≤ 1 we have σ(U) ⊆ ¦z ∈ C[ [z[ ≤ 1¦. Moreover, U
−1
is also unitary and hence σ(U) ⊆ ¦z ∈ C[ [z[ ≥ 1¦ by Lemma 2.13. If
Uψ
j
= z
j
ψ
j
, j = 1, 2 we have
(z
1
−z
2
)'ψ
1
, ψ
2
` = 'U
∗
ψ
1
, ψ
2
` −'ψ
1
, Uψ
2
` = 0 (2.100)
since Uψ = zψ implies U
∗
ψ = U
−1
ψ = z
−1
ψ = z
∗
ψ.
Problem 2.10. What is the spectrum of an orthogonal projection?
Problem 2.11. Compute the resolvent of Af = f
, D(A) = ¦f ∈ H
1
[0, 1] [ f(0) =
0¦ and show that unbounded operators can have empty spectrum.
Problem 2.12. Compute the eigenvalues and eigenvectors of A = −
d
2
dx
2
,
D(A) = ¦f ∈ H
2
(0, π)[f(0) = f(π) = 0¦. Compute the resolvent of A.
2.4. Orthogonal sums of operators 67
Problem 2.13. Find a Weyl sequence for the selfadjoint operator A =
−
d
2
dx
2
, D(A) = H
2
(R) for z ∈ (0, ∞). What is σ(A)? (Hint: Cut oﬀ the
solutions of −u
(x) = z u(x) outside a ﬁnite ball.)
Problem 2.14. Suppose A is bounded. Show that the spectrum of AA
∗
and
A
∗
A coincide away from 0 by showing
R
AA
∗(z) =
1
z
(AR
A
∗
A
(z)A
∗
−1) , R
A
∗
A
(z) =
1
z
(A
∗
R
AA
∗(z)A−1) .
(2.101)
2.4. Orthogonal sums of operators
Let H
j
, j = 1, 2, be two given Hilbert spaces and let A
j
: D(A
j
) → H
j
be
two given operators. Setting H = H
1
⊕H
2
we can deﬁne an operator
A = A
1
⊕A
2
, D(A) = D(A
1
) ⊕D(A
2
) (2.102)
by setting A(ψ
1
+ψ
2
) = A
1
ψ
1
+A
2
ψ
2
for ψ
j
∈ D(A
j
). Clearly A is closed,
(essentially) selfadjoint, etc., if and only if both A
1
and A
2
are. The same
considerations apply to countable orthogonal sums
A =
j
A
j
, D(A) =
j
D(A
j
) (2.103)
and we have
Theorem 2.20. Suppose A
j
are selfadjoint operators on H
j
, then A =
¸
j
A
j
is selfadjoint and
R
A
(z) =
j
R
A
j
(z), z ∈ ρ(A) = C`σ(A) (2.104)
where
σ(A) =
¸
j
σ(A
j
) (2.105)
(the closure can be omitted if there are only ﬁnitely many terms).
Proof. By Ran(A±i) = (A±i)D(A) =
¸
j
(A
j
±i)D(A
j
) =
¸
j
H
j
= H we
see that A is selfadjoint. Moreover, if z ∈ σ(A
j
) there is a corresponding
Weyl sequence ψ
n
∈ D(A
j
) ⊆ D(A) and hence z ∈ σ(A). Conversely, if
z ∈
¸
j
σ(A
j
) set ε = d(z,
¸
j
σ(A
j
)) > 0, then R
A
j
(z) ≤ ε
−1
and hence

¸
j
R
A
j
(z) ≤ ε
−1
shows that z ∈ ρ(A).
Conversely, given an operator A it might be useful to write A as orthog
onal sum and investigate each part separately.
Let H
1
⊆ H be a closed subspace and let P
1
be the corresponding projec
tor. We say that H
1
reduces the operator A if P
1
A ⊆ AP
1
. Note that this
68 2. Selfadjointness and spectrum
implies P
1
D(A) ⊆ D(A). Moreover, if we set H
2
= H
⊥
1
, we have H = H
1
⊕H
2
and P
2
= I −P
1
reduces A as well.
Lemma 2.21. Suppose H
1
⊆ H reduces A, then A = A
1
⊕A
2
, where
A
j
ψ = Aψ, D(A
j
) = P
j
D(A) ⊆ D(A). (2.106)
If A is closable, then H
1
also reduces A and
A = A
1
⊕A
2
. (2.107)
Proof. As already noted, P
1
D(A) ⊆ D(A) and hence P
2
D(A) = (I −
P
1
)D(A) ⊆ D(A). Thus we see D(A) = D(A
1
) ⊕ D(A
2
). Moreover, if
ψ ∈ D(A
j
) we have Aψ = AP
j
ψ = P
j
Aψ ∈ H
j
and thus A
j
: D(A
j
) → H
j
which proves the ﬁrst claim.
Now let us turn to the second claim. Clearly A ⊆ A
1
⊕A
2
. Conversely,
suppose ψ ∈ D(A), then there is a sequence ψ
n
∈ D(A) such that ψ
n
→ ψ
and Aψ
n
→ Aψ. Then P
j
ψ
n
→ P
j
ψ and AP
j
ψ
n
= P
j
Aψ
n
→ PAψ. In
particular, P
j
ψ ∈ D(A) and AP
j
ψ = PAψ.
If A is selfadjoint, then H
1
reduces A if P
1
D(A) ⊆ D(A) and AP
1
ψ ∈ H
1
for every ψ ∈ D(A). In fact, if ψ ∈ D(A) we can write ψ = ψ
1
⊕ ψ
2
,
ψ
j
= P
j
ψ ∈ D(A). Since AP
1
ψ = Aψ
1
and P
1
Aψ = P
1
Aψ
1
+ P
1
Aψ
2
=
Aψ
1
+P
1
Aψ
2
we need to show P
1
Aψ
2
= 0. But this follows since
'ϕ, P
1
Aψ
2
` = 'AP
1
ϕ, ψ
2
` = 0 (2.108)
for every ϕ ∈ D(A).
Problem 2.15. Show (A
1
⊕A
2
)
∗
= A
∗
1
⊕A
∗
2
.
2.5. Selfadjoint extensions
It is safe to skip this entire section on ﬁrst reading.
In many physical applications a symmetric operator is given. If this
operator turns out to be essentially selfadjoint, there is a unique selfadjoint
extension and everything is ﬁne. However, if it is not, it is important to ﬁnd
out if there are selfadjoint extensions at all (for physical problems there
better are) and to classify them.
In Section 2.2 we have seen that A is essentially selfadjoint if Ker(A
∗
−
z) = Ker(A
∗
−z
∗
) = ¦0¦ for one z ∈ C`R. Hence selfadjointness is related
to the dimension of these spaces and one calls the numbers
d
±
(A) = dimK
±
, K
±
= Ran(A±i)
⊥
= Ker(A
∗
∓i), (2.109)
defect indices of A (we have chosen z = i for simplicity, any other z ∈ C`R
would be as good). If d
−
(A) = d
+
(A) = 0 there is one selfadjoint extension
of A, namely A. But what happens in the general case? Is there more than
2.5. Selfadjoint extensions 69
one extension, or maybe none at all? These questions can be answered by
virtue of the Cayley transform
V = (A−i)(A+ i)
−1
: Ran(A+ i) → Ran(A−i). (2.110)
Theorem 2.22. The Cayley transform is a bijection from the set of all
symmetric operators A to the set of all isometric operators V (i.e., V ϕ =
ϕ for all ϕ ∈ D(V )) for which Ran(1 +V ) is dense.
Proof. Since A is symmetric we have (A ± i)ψ
2
= Aψ
2
+ ψ
2
for
all ψ ∈ D(A) by a straightforward computation. And thus for every ϕ =
(A+ i)ψ ∈ D(V ) = Ran(A+ i) we have
V ϕ = (A−i)ψ = (A+ i)ψ = ϕ. (2.111)
Next observe
1 ±V = ((A−i) ±(A+ i))(A+ i)
−1
=
2A(A+ i)
−1
2i(A+ i)
−1
, (2.112)
which shows D(A) = Ran(1 −V ) and
A = i(1 +V )(1 −V )
−1
. (2.113)
Conversely, let V be given and use the last equation to deﬁne A.
Since V is isometric we have '(1 ± V )ϕ, (1 ∓ V )ϕ` = ±2iIm'V ϕ, ϕ`
for all ϕ ∈ D(V ) by a straightforward computation. And thus for every
ψ = (1 −V )ϕ ∈ D(A) = Ran(1 −V ) we have
'Aψ, ψ` = −i'(1+V )ϕ, (1−V )ϕ` = i'(1−V )ϕ, (1+V )ϕ` = 'ψ, Aψ`, (2.114)
that is, A is symmetric. Finally observe
A±i = ((1 +V ) ±(1 −V ))(1 −V )
−1
=
2i(1 −V )
−1
2iV (1 −V )
−1
, (2.115)
which shows that A is the Cayley transform of V and ﬁnishes the proof.
Thus A is selfadjoint if and only if its Cayley transform V is unitary.
Moreover, ﬁnding a selfadjoint extension of A is equivalent to ﬁnding a
unitary extensions of V and this in turn is equivalent to (taking the closure
and) ﬁnding a unitary operator from D(V )
⊥
to Ran(V )
⊥
. This is possible
if and only if both spaces have the same dimension, that is, if and only if
d
+
(A) = d
−
(A).
Theorem 2.23. A symmetric operator has selfadjoint extensions if and
only if its defect indices are equal.
In this case let A
1
be a selfadjoint extension, V
1
its Cayley transform.
Then
D(A
1
) = D(A) + (1 −V
1
)K
+
= ¦ψ +ϕ
+
−V
1
ϕ
+
[ψ ∈ D(A), ϕ
+
∈ K
+
¦
(2.116)
70 2. Selfadjointness and spectrum
and
A
1
(ψ +ϕ
+
−V
1
ϕ
+
) = Aψ + iϕ
+
+ iV
1
ϕ
+
. (2.117)
Moreover,
(A
1
±i)
−1
= (A±i)
−1
⊕
∓i
2
¸
j
'ϕ
±
j
, .`(ϕ
±
j
−ϕ
∓
j
), (2.118)
where ¦ϕ
+
j
¦ is an orthonormal basis for K
+
and ϕ
−
j
= V
1
ϕ
+
j
.
Corollary 2.24. Suppose A is a closed symmetric operator with equal defect
indices d = d
+
(A) = d
−
(A). Then dimKer(A
∗
−z
∗
) = d for all z ∈ C`R.
Proof. First of all we note that instead of z = i we could use V (z) =
(A + z
∗
)(A + z)
−1
for any z ∈ C`R. Let d
±
(z) = dimK
±
(z), K
+
(z) =
Ran(A + z)
⊥
respectively K
−
(z) = Ran(A + z
∗
)
⊥
. The same arguments
as before show that there is a one to one correspondence between the self
adjoint extensions of A and the unitary operators on C
d(z)
. Hence d(z
1
) =
d(z
2
) = d
±
(A).
Example. Recall the operator A = −i
d
dx
, D(A) = ¦f ∈ H
1
(0, 2π)[f(0) =
f(2π) = 0¦ with adjoint A
∗
= −i
d
dx
, D(A
∗
) = H
1
(0, 2π).
Clearly
K
±
= span¦e
∓x
¦ (2.119)
is one dimensional and hence all unitary maps are of the form
V
θ
e
2π−x
= e
iθ
e
x
, θ ∈ [0, 2π). (2.120)
The functions in the domain of the corresponding operator A
θ
are given by
f
θ
(x) = f(x) +α(e
2π−x
−e
iθ
e
x
), f ∈ D(A), α ∈ C. (2.121)
In particular, f
θ
satisﬁes
f
θ
(2π) = e
i
˜
θ
f
θ
(0), e
i
˜
θ
=
1 −e
iθ
e
2π
e
2π
−e
iθ
(2.122)
and thus we have
D(A
θ
) = ¦f ∈ H
1
(0, 2π)[f(2π) = e
i
˜
θ
f(0)¦. (2.123)
Concerning closures we note that a bounded operator is closed if and
only if its domain is closed and any operator is closed if and only if its
inverse is closed. Hence we have
Lemma 2.25. The following items are equivalent.
• A is closed.
• D(V ) = Ran(A+ i) is closed.
2.5. Selfadjoint extensions 71
• Ran(V ) = Ran(A−i) is closed.
• V is closed.
Next, we give a useful criterion for the existence of selfadjoint exten
sions. A skew linear map C : H → H is called a conjugation if it satisﬁes
C
2
= I and 'Cψ, Cϕ` = 'ψ, ϕ`. The prototypical example is of course
complex conjugation Cψ = ψ
∗
. An operator A is called Creal if
CD(A) ⊆ D(A), and ACψ = CAψ, ψ ∈ D(A). (2.124)
Note that in this case CD(A) = D(A), since D(A) = C
2
D(A) ⊆ CD(A).
Theorem 2.26. Suppose the symmetric operator A is Creal, then its defect
indices are equal.
Proof. Let ¦ϕ
j
¦ be an orthonormal set in Ran(A+i)
⊥
. Then ¦Kϕ
j
¦ is an
orthonormal set in Ran(A − i)
⊥
. Hence ¦ϕ
j
¦ is an orthonormal basis for
Ran(A+ i)
⊥
if and only if ¦Kϕ
j
¦ is an orthonormal basis for Ran(A−i)
⊥
.
Hence the two spaces have the same dimension.
Finally, we note the following useful formula for the diﬀerence of resol
vents of selfadjoint extensions.
Lemma 2.27. If A
j
, j = 1, 2 are selfadjoint extensions and if ¦ϕ
j
(z)¦ is
an orthonormal basis for Ker(A
∗
−z
∗
), then
(A
1
−z)
−1
−(A
2
−z)
−1
=
¸
j,k
(α
1
jk
(z) −α
2
jk
(z))'ϕ
k
(z), .`ϕ
k
(z
∗
), (2.125)
where
α
l
jk
(z) = 'ϕ
j
(z
∗
), (A
l
−z)
−1
ϕ
k
(z)`. (2.126)
Proof. First observe that ((A
1
−z)
−1
−(A
2
−z)
−1
)ϕ is zero for every ϕ ∈
Ran(A−z). Hence it suﬃces to consider it for vectors ϕ =
¸
j
'ϕ
j
(z), ϕ`ϕ
j
(z) ∈
Ran(A−z)
⊥
. Hence we have
(A
1
−z)
−1
−(A
2
−z)
−1
=
¸
j
'ϕ
j
(z), .`ψ
j
(z), (2.127)
where
ψ
j
(z) = ((A
1
−z)
−1
−(A
2
−z)
−1
)ϕ
j
(z). (2.128)
Now computation the adjoint once using ((A
j
− z)
−1
)
∗
= (A
j
− z
∗
)
−1
and
once using (
¸
j
'ψ
j
, .`ϕ
j
)
∗
=
¸
j
'ϕ
j
, .`ψ
j
we obtain
¸
j
'ϕ
j
(z
∗
), .`ψ
j
(z
∗
) =
¸
j
'ψ
j
(z), .`ϕ
j
(z). (2.129)
72 2. Selfadjointness and spectrum
Evaluating at ϕ
k
(z) implies
ψ
k
(z) =
¸
j
'ψ
j
(z
∗
), ϕ
k
(z)`ϕ
j
(z
∗
) (2.130)
and ﬁnishes the proof.
Problem 2.16. Compute the defect indices of A
0
= i
d
dx
, D(A
0
) = C
∞
c
((0, ∞)).
Can you give a selfadjoint extension of A
0
.
2.6. Appendix: Absolutely continuous functions
Let (a, b) ⊆ R be some interval. We denote by
AC(a, b) = ¦f ∈ C(a, b)[f(x) = f(c) +
x
c
g(t)dt, c ∈ (a, b), g ∈ L
1
loc
(a, b)¦
(2.131)
the set of all absolutely continuous functions. That is, f is absolutely
continuous if and only if it can be written as the integral of some locally
integrable function. Note that AC(a, b) is a vector space.
By Corollary A.33 f(x) = f(c) +
x
c
g(t)dt is diﬀerentiable a.e. (with re
spect to Lebesgue measure) and f
(x) = g(x). In particular, g is determined
uniquely a.e..
If [a, b] is a compact interval we set
AC[a, b] = ¦f ∈ AC(a, b)[g ∈ L
1
(a, b)¦ ⊆ C[a, b]. (2.132)
If f, g ∈ AC[a, b] we have the formula of partial integration
b
a
f(x)g
(x) = f(b)g(b) −f(a)g(a) −
b
a
f
(x)g(x)dx. (2.133)
We set
H
m
(a, b) = ¦f ∈ L
2
(a, b)[f
(j)
∈ AC(a, b), f
(j+1)
∈ L
2
(a, b), 0 ≤ j ≤ m−1¦.
(2.134)
Then we have
Lemma 2.28. Suppose f ∈ H
m
(a, b), m ≥ 1. Then f is bounded and
lim
x↓a
f
(j)
(x) respectively lim
x↑b
f
(j)
(x) exist for 0 ≤ j ≤ m−1. Moreover,
the limit is zero if the endpoint is inﬁnite.
Proof. If the endpoint is ﬁnite, then f
(j+1)
is integrable near this endpoint
and hence the claim follows. If the endpoint is inﬁnite, note that
[f
(j)
(x)[
2
= [f
(j)
(c)[
2
+ 2
x
c
Re(f
(j)
(t)
∗
f
(j+1)
(t))dt (2.135)
shows that the limit exists (dominated convergence). Since f
(j)
is square
integrable the limit must be zero.
2.6. Appendix: Absolutely continuous functions 73
Let me remark, that it suﬃces to check that the function plus the highest
derivative is in L
2
, the lower derivatives are then automatically in L
2
. That
is,
H
m
(a, b) = ¦f ∈ L
2
(a, b)[f
(j)
∈ AC(a, b), 0 ≤ j ≤ m−1, f
(r)
∈ L
2
(a, b)¦.
(2.136)
For a ﬁnite endpoint this is straightforward. For an inﬁnite endpoint this
can also be shown directly, but it is much easier to use the Fourier transform
(compare Section 7.1).
Problem 2.17. Show (2.133). (Hint: Fubini)
Problem 2.18. Show that H
1
(a, b) together with the norm
f
2
2,1
=
b
a
[f(t)[
2
dt +
b
a
[f
(t)[
2
dt (2.137)
is a Hilbert space.
Problem 2.19. What is the closure of C
∞
0
(a, b) in H
1
(a, b)? (Hint: Start
with the case where (a, b) is ﬁnite.)
Chapter 3
The spectral theorem
The time evolution of a quantum mechanical system is governed by the
Schr¨odinger equation
i
d
dt
ψ(t) = Hψ(t). (3.1)
If H = C
n
, and H is hence a matrix, this system of ordinary diﬀerential
equations is solved by the matrix exponential
ψ(t) = exp(−itH)ψ(0). (3.2)
This matrix exponential can be deﬁned by a convergent power series
exp(−itH) =
∞
¸
n=0
(−it)
n
n!
H
n
. (3.3)
For this approach the boundedness of H is crucial, which might not be the
case for a a quantum system. However, the best way to compute the matrix
exponential, and to understand the underlying dynamics, is to diagonalize
H. But how do we diagonalize a selfadjoint operator? The answer is known
as the spectral theorem.
3.1. The spectral theorem
In this section we want to address the problem of deﬁning functions of a
selfadjoint operator A in a natural way, that is, such that
(f+g)(A) = f(A)+g(A), (fg)(A) = f(A)g(A), (f
∗
)(A) = f(A)
∗
. (3.4)
As long as f and g are polynomials, no problems arise. If we want to extend
this deﬁnition to a larger class of functions, we will need to perform some
limiting procedure. Hence we could consider convergent power series or
equip the space of polynomials with the sup norm. In both cases this only
75
76 3. The spectral theorem
works if the operator A is bounded. To overcome this limitation, we will use
characteristic functions χ
Ω
(A) instead of powers A
j
. Since χ
Ω
(λ)
2
= χ
Ω
(λ),
the corresponding operators should be orthogonal projections. Moreover,
we should also have χ
R
(A) = I and χ
Ω
(A) =
¸
n
j=1
χ
Ω
j
(A) for any ﬁnite
union Ω =
¸
n
j=1
Ω
j
of disjoint sets. The only remaining problem is of course
the deﬁnition of χ
Ω
(A). However, we will defer this problem and begin
by developing a functional calculus for a family of characteristic functions
χ
Ω
(A).
Denote the Borel sigma algebra of R by B. A projectionvalued mea
sure is a map
P : B →L(H), Ω → P(Ω), (3.5)
from the Borel sets to the set of orthogonal projections, that is, P(Ω)
∗
=
P(Ω) and P(Ω)
2
= P(Ω), such that the following two conditions hold:
(i) P(R) = I.
(ii) If Ω =
¸
n
Ω
n
with Ω
n
∩ Ω
m
= ∅ for n = m, then
¸
n
P(Ω
n
)ψ =
P(Ω)ψ for every ψ ∈ H (strong σadditivity).
Note that we require strong convergence,
¸
n
P(Ω
n
)ψ = P(Ω)ψ, rather
than norm convergence,
¸
n
P(Ω
n
) = P(Ω). In fact, norm convergence
does not even hold in the simplest case where H = L
2
(I) and P(Ω) = χ
Ω
(multiplication operator), since for a multiplication operator the norm is just
the sup norm of the function. Furthermore, it even suﬃces to require weak
convergence, since wlimP
n
= P for some orthogonal projections implies
slimP
n
= P by 'ψ, P
n
ψ` = 'ψ, P
2
n
ψ` = 'P
n
ψ, P
n
ψ` = P
n
ψ
2
together
with Lemma 1.11 (iv).
Example. Let H = C
n
and A ∈ GL(n) be some symmetric matrix. Let
λ
1
, . . . , λ
m
be its (distinct) eigenvalues and let P
j
be the projections onto
the corresponding eigenspaces. Then
P
A
(Ω) =
¸
{jλ
j
∈Ω}
P
j
(3.6)
is a projection valued measure.
Example. Let H = L
2
(R) and let f be a realvalued measurable function.
Then
P(Ω) = χ
f
−1
(Ω)
(3.7)
is a projection valued measure (Problem 3.2).
It is straightforward to verify that any projectionvalued measure satis
ﬁes
P(∅) = 0, P(R`Ω) = I −P(Ω), (3.8)
3.1. The spectral theorem 77
and
P(Ω
1
∪ Ω
2
) +P(Ω
1
∩ Ω
2
) = P(Ω
1
) +P(Ω
2
). (3.9)
Moreover, we also have
P(Ω
1
)P(Ω
2
) = P(Ω
1
∩ Ω
2
). (3.10)
Indeed, suppose Ω
1
∩Ω
2
= ∅ ﬁrst. Then, taking the square of (3.9) we infer
P(Ω
1
)P(Ω
2
) +P(Ω
2
)P(Ω
1
) = 0. (3.11)
Multiplying this equation from the right by P(Ω
2
) shows that P(Ω
1
)P(Ω
2
) =
−P(Ω
2
)P(Ω
1
)P(Ω
2
) is selfadjoint and thus P(Ω
1
)P(Ω
2
) = P(Ω
2
)P(Ω
1
) =
0. For the general case Ω
1
∩ Ω
2
= ∅ we now have
P(Ω
1
)P(Ω
2
) = (P(Ω
1
−Ω
2
) +P(Ω
1
∩ Ω
2
))(P(Ω
2
−Ω
1
) +P(Ω
1
∩ Ω
2
))
= P(Ω
1
∩ Ω
2
) (3.12)
as stated.
To every projectionvalued measure there corresponds a resolution of
the identity
P(λ) = P((−∞, λ]) (3.13)
which has the properties (Problem 3.3):
(i) P(λ) is an orthogonal projection.
(ii) P(λ
1
) ≤ P(λ
2
) (that is 'ψ, P(λ
1
)ψ` ≤ 'ψ, P(λ
2
)ψ`) or equiva
lently Ran(P(λ
1
)) ⊆ Ran(P(λ
2
)) for λ
1
≤ λ
2
.
(iii) slim
λn↓λ
P(λ
n
) = P(λ) (strong right continuity).
(iv) slim
λ→−∞
P(λ) = 0 and slim
λ→+∞
P(λ) = I.
As before, strong right continuity is equivalent to weak right continuity.
Picking ψ ∈ H, we obtain a ﬁnite Borel measure µ
ψ
(Ω) = 'ψ, P(Ω)ψ` =
P(Ω)ψ
2
with µ
ψ
(R) = ψ
2
< ∞. The corresponding distribution func
tion is given by µ(λ) = 'ψ, P(λ)ψ` and since for every distribution function
there is a unique Borel measure (Theorem A.2), for every resolution of the
identity there is a unique projection valued measure.
Using the polarization identity (2.16) we also have the following complex
Borel measures
µ
ϕ,ψ
(Ω) = 'ϕ, P(Ω)ψ` =
1
4
(µ
ϕ+ψ
(Ω) −µ
ϕ−ψ
(Ω) + iµ
ϕ−iψ
(Ω) −iµ
ϕ+iψ
(Ω)).
(3.14)
Note also that, by CauchySchwarz, [µ
ϕ,ψ
(Ω)[ ≤ ϕ ψ.
Now let us turn to integration with respect to our projectionvalued
measure. For any simple function f =
¸
n
j=1
α
j
χ
Ω
j
(where Ω
j
= f
−1
(α
j
))
78 3. The spectral theorem
we set
P(f) ≡
R
f(λ)dP(λ) =
n
¸
j=1
α
j
P(Ω
j
). (3.15)
In particular, P(χ
Ω
) = P(Ω). Then 'ϕ, P(f)ψ` =
¸
j
α
j
µ
ϕ,ψ
(Ω
j
) shows
'ϕ, P(f)ψ` =
R
f(λ)dµ
ϕ,ψ
(λ) (3.16)
and, by linearity of the integral, the operator P is a linear map from the set
of simple functions into the set of bounded linear operators on H. Moreover,
P(f)ψ
2
=
¸
j
[α
j
[
2
µ
ψ
(Ω
j
) (the sets Ω
j
are disjoint) shows
P(f)ψ
2
=
R
[f(λ)[
2
dµ
ψ
(λ). (3.17)
Equipping the set of simple functions with the sup norm this implies
P(f)ψ ≤ f
∞
ψ (3.18)
that P has norm one. Since the simple functions are dense in the Banach
space of bounded Borel functions B(R), there is a unique extension of P to
a bounded linear operator P : B(R) → L(H) (whose norm is one) from the
bounded Borel functions on R (with sup norm) to the set of bounded linear
operators on H. In particular, (3.16) and (3.17) remain true.
There is some additional structure behind this extension. Recall that
the set L(H) of all bounded linear mappings on H forms a C
∗
algebra. A C
∗
algebra homomorphism φ is a linear map between two C
∗
algebras which
respects both the multiplication and the adjoint, that is, φ(ab) = φ(a)φ(b)
and φ(a
∗
) = φ(a)
∗
.
Theorem 3.1. Let P(Ω) be a projectionvalued measure on H. Then the
operator
P : B(R) → L(H)
f →
R
f(λ)dP(λ)
(3.19)
is a C
∗
algebra homomorphism with norm one.
In addition, if f
n
(x) → f(x) pointwise and if the sequence sup
λ∈R
[f
n
(λ)[
is bounded, then P(f
n
) → P(f) strongly.
Proof. The properties P(1) = I, P(f
∗
) = P(f)
∗
, and P(fg) = P(f)P(g)
are straightforward for simple functions f. For general f they follow from
continuity. Hence P is a C
∗
algebra homomorphism.
The last claim follows from the dominated convergence theorem and
(3.17).
3.1. The spectral theorem 79
As a consequence, observe
'P(g)ϕ, P(f)ψ` =
R
g
∗
(λ)f(λ)dµ
ϕ,ψ
(λ) (3.20)
and thus
dµ
P(g)ϕ,P(f)ψ
= g
∗
fdµ
ϕ,ψ
. (3.21)
Example. Let H = C
n
and A ∈ GL(n) respectively P
A
as in the previous
example. Then
P
A
(f) =
m
¸
j=1
f(λ
j
)P
j
. (3.22)
In particular, P
A
(f) = A for f(λ) = λ.
Next we want to deﬁne this operator for unbounded Borel functions.
Since we expect the resulting operator to be unbounded, we need a suitable
domain ﬁrst. Motivated by (3.17) we set
D
f
= ¦ψ ∈ H[
R
[f(λ)[
2
dµ
ψ
(λ) < ∞¦. (3.23)
This is clearly a linear subspace of H since µ
αψ
(Ω) = [α[
2
µ
ψ
(Ω) and since
µ
ϕ+ψ
(Ω) ≤ 2(µ
ϕ
(Ω) +µ
ψ
(Ω)) (by the triangle inequality).
For every ψ ∈ D
f
, the bounded Borel function
f
n
= χ
Ωn
f, Ω
n
= ¦λ[ [f(λ)[ ≤ n¦, (3.24)
converges to f in the sense of L
2
(R, dµ
ψ
). Moreover, because of (3.17),
P(f
n
)ψ converges to some vector
˜
ψ. We deﬁne P(f)ψ =
˜
ψ. By construction,
P(f) is a linear operator such that (3.16) and (3.17) hold.
In addition, D
f
is dense. Indeed, let Ω
n
be deﬁned as in (3.24) and
abbreviate ψ
n
= P(Ω
n
)ψ. Now observe that dµ
ψn
= χ
Ωn
dµ
ψ
and hence
ψ
n
∈ D
f
. Moreover, ψ
n
→ ψ by (3.17) since χ
Ωn
→ 1 in L
2
(R, dµ
ψ
).
The operator P(f) has some additional properties. One calls an un
bounded operator A normal if D(A) = D(A
∗
) and Aψ = A
∗
ψ for all
ψ ∈ D(A).
Theorem 3.2. For any Borel function f, the operator
P(f) =
R
f(λ)dP(λ), D(P(f)) = D
f
, (3.25)
is normal and satisﬁes
P(f)
∗
= P(f
∗
). (3.26)
Proof. Let f be given and deﬁne f
n
, Ω
n
as above. Since (3.26) holds for
f
n
by our previous theorem, we get
'ϕ, P(f)ψ` = 'P(f
∗
)ϕ, ψ` (3.27)
80 3. The spectral theorem
for any ϕ, ψ ∈ D
f
= D(f
∗
) by continuity. Thus it remains to show that
D(P(f)
∗
) ⊆ D
f
. If ψ ∈ D(P(f)
∗
) we have 'ψ, P(f)ϕ` = '
˜
ψ, ϕ` for all
ϕ ∈ D
f
by deﬁnition. Now observe that P(f
∗
n
)ψ = P(Ω
n
)
˜
ψ since we have
'P(f
∗
n
)ψ, ϕ` = 'ψ, P(f
n
)ϕ` = 'ψ, P(f)P(Ω
n
)ϕ` = 'P(Ω
n
)
˜
ψ, ϕ` (3.28)
for any ϕ ∈ H. To see the second equality use P(f
n
)ϕ = P(f
m
χ
n
)ϕ =
P(f
m
)P(Ω
n
)ϕ for m ≥ n and let m → ∞. This proves existence of the limit
lim
n→∞
R
[f
n
[
2
dµ
ψ
(λ) = lim
n→∞
P(f
∗
n
)ψ
2
= lim
n→∞
P(Ω
n
)
˜
ψ
2
= 
˜
ψ
2
, (3.29)
which implies f ∈ L
2
(R, dµ
ψ
), that is, ψ ∈ D
f
. That P(f) is normal follows
from P(f)ψ
2
= P(f
∗
)ψ
2
=
R
[f(λ)[
2
dµ
ψ
.
These considerations seem to indicate some kind of correspondence be
tween the operators P(f) in H and f in L
2
(R, dµ
ψ
). Recall that U : H →
˜
H is
called unitary if it is a bijection which preserves scalar products 'Uϕ, Uψ` =
'ϕ, ψ`. The operators A in H and
˜
A in
˜
H are said to be unitarily equiva
lent if
UA =
˜
AU, UD(A) = D(
˜
A). (3.30)
Clearly, A is selfadjoint if and only if
˜
A is and σ(A) = σ(
˜
A).
Now let us return to our original problem and consider the subspace
H
ψ
= ¦P(f)ψ[f ∈ L
2
(R, dµ
ψ
)¦ ⊆ H. (3.31)
Observe that this subspace is closed: If ψ
n
= P(f
n
)ψ converges in H, then
f
n
converges to some f in L
2
(since ψ
n
− ψ
m

2
=
[f
n
− f
m
[
2
dµ
ψ
) and
hence ψ
n
→ P(f)ψ.
The vector ψ is called cyclic if H
ψ
= H. By (3.17), the relation
U
ψ
(P(f)ψ) = f (3.32)
deﬁnes a unique unitary operator U
ψ
: H
ψ
→ L
2
(R, dµ
ψ
) such that
U
ψ
P(f) = fU
ψ
, (3.33)
where f is identiﬁed with its corresponding multiplication operator. More
over, if f is unbounded we have U
ψ
(D
f
∩H
ψ
) = D(f) = ¦g ∈ L
2
(R, dµ
ψ
)[fg ∈
L
2
(R, dµ
ψ
)¦ (since ϕ = P(f)ψ implies dµ
ϕ
= [f[
2
dµ
ψ
) and the above equa
tion still holds.
If ψ is cyclic, our picture is complete. Otherwise we need to extend this
approach. A set ¦ψ
j
¦
j∈J
(J some index set) is called a set of spectral vectors
if ψ
j
 = 1 and H
ψ
i
⊥ H
ψ
j
for all i = j. A set of spectral vectors is called a
spectral basis if
¸
j
H
ψ
j
= H. Luckily a spectral basis always exist:
3.1. The spectral theorem 81
Lemma 3.3. For every projection valued measure P, there is an (at most
countable) spectral basis ¦ψ
n
¦ such that
H =
n
H
ψn
(3.34)
and a corresponding unitary operator
U =
n
U
ψn
: H →
n
L
2
(R, dµ
ψn
) (3.35)
such that for any Borel function f,
UP(f) = fU, UD
f
= D(f). (3.36)
Proof. It suﬃces to show that a spectral basis exists. This can be easily
done using a GramSchmidt type construction. First of all observe that if
¦ψ
j
¦
j∈J
is a spectral set and ψ ⊥ H
ψ
j
for all j we have H
ψ
⊥ H
ψ
j
for all j.
Indeed, ψ ⊥ H
ψ
j
implies P(g)ψ ⊥ H
ψ
j
for every bounded function g since
'P(g)ψ, P(f)ψ
j
` = 'ψ, P(g
∗
f)ψ
j
` = 0. But P(g)ψ with g bounded is dense
in H
ψ
implying H
ψ
⊥ H
ψ
j
.
Now start with some total set ¦
˜
ψ
j
¦. Normalize
˜
ψ
1
and choose this to be
ψ
1
. Move to the ﬁrst
˜
ψ
j
which is not in H
ψ
1
, take the orthogonal complement
with respect to H
ψ
1
and normalize it. Choose the result to be ψ
2
. Proceeding
like this we get a set of spectral vectors ¦ψ
j
¦ such that span¦
˜
ψ
j
¦ ⊆
¸
j
H
ψ
j
.
Hence H = span¦
˜
ψ
j
¦ ⊆
¸
j
H
ψ
j
.
It is important to observe that the cardinality of a spectral basis is not
welldeﬁned (in contradistinction to the cardinality of an ordinary basis of
the Hilbert space). However, it can be at most equal to the cardinality of an
ordinary basis. In particular, since H is separable, it is at most countable.
The minimal cardinality of a spectral basis is called spectral multiplicity
of P. If the spectral multiplicity is one, the spectrum is called simple.
Example. Let H = C
2
and A =
0 0
0 1
. Then ψ
1
= (1, 0) and ψ
2
=
(0, 1) are a spectral basis. However, ψ = (1, 1) is cyclic and hence the
spectrum of A is simple. If A =
1 0
0 1
there is no cyclic vector (why)
and hence the spectral multiplicity is two.
Using this canonical form of projection valued measures it is straight
forward to prove
Lemma 3.4. Let f, g be Borel functions and α, β ∈ C. Then we have
αP(f) +βP(g) ⊆ P(αf +βg), D(αP(f) +βP(g)) = D
f+g
(3.37)
82 3. The spectral theorem
and
P(f)P(g) ⊆ P(f g), D(P(f)P(g)) = D
g
∩ D
f g
. (3.38)
Now observe, that to every projection valued measure P we can assign a
selfadjoint operator A =
R
λdP(λ). The question is whether we can invert
this map. To do this, we consider the resolvent R
A
(z) =
R
(λ −z)
−1
dP(λ).
By (3.16) the corresponding quadratic form is given by
F
ψ
(z) = 'ψ, R
A
(z)ψ` =
R
1
λ −z
dµ
ψ
(λ), (3.39)
which is know as the Borel transform of the measure µ
ψ
. It can be shown
(see Section 3.4) that F
ψ
(z) is a holomorphic map from the upper half
plane to itself. Such functions are called Herglotz functions. Moreover,
the measure µ
ψ
can be reconstructed from F
ψ
(z) by Stieltjes inversion
formula
µ
ψ
(λ) = lim
δ↓0
lim
ε↓0
1
π
λ+δ
−∞
Im(F
ψ
(t + iε))dt. (3.40)
Conversely, if F
ψ
(z) is a Herglotz function satisfying [F(z)[ ≤
M
Im(z)
, then it
is the Borel transform of a unique measure µ
ψ
(given by Stieltjes inversion
formula).
So let A be a given selfadjoint operator and consider the expectation of
the resolvent of A,
F
ψ
(z) = 'ψ, R
A
(z)ψ`. (3.41)
This function is holomorphic for z ∈ ρ(A) and satisﬁes
F
ψ
(z
∗
) = F
ψ
(z)
∗
and [F
ψ
(z)[ ≤
ψ
2
Im(z)
(3.42)
(see Theorem 2.14). Moreover, the ﬁrst resolvent formula (2.80) shows
Im(F
ψ
(z)) = Im(z)R
A
(z)ψ
2
(3.43)
that it maps the upper half plane to itself, that is, it is a Herglotz function.
So by our above remarks, there is a corresponding measure µ
ψ
(λ) given by
Stieltjes inversion formula. It is called spectral measure corresponding to
ψ.
More generally, by polarization, for each ϕ, ψ ∈ H we can ﬁnd a corre
sponding complex measure µ
ϕ,ψ
such that
'ϕ, R
A
(z)ψ` =
R
1
λ −z
dµ
ϕ,ψ
(λ). (3.44)
The measure µ
ϕ,ψ
is conjugate linear in ϕ and linear in ψ. Moreover, a
comparison with our previous considerations begs us to deﬁne a family of
3.1. The spectral theorem 83
operators P
A
(Ω) via
'ϕ, P
A
(Ω)ψ` =
R
χ
Ω
(λ)dµ
ϕ,ψ
(λ). (3.45)
This is indeed possible by Corollary 1.8 since ['ϕ, P
A
(Ω)ψ`[ = [µ
ϕ,ψ
(Ω)[ ≤
ϕ ψ. The operators P
A
(Ω) are non negative (0 ≤ 'ψ, P
A
(Ω)ψ` ≤ 1) and
hence selfadjoint.
Lemma 3.5. The family of operators P
A
(Ω) forms a projection valued mea
sure.
Proof. We ﬁrst show P
A
(Ω
1
)P
A
(Ω
2
) = P
A
(Ω
1
∩ Ω
2
) in two steps. First
observe (using the ﬁrst resolvent formula (2.80))
R
1
λ − ˜ z
dµ
R
A
(z
∗
)ϕ,ψ
(λ) = 'R
A
(z
∗
)ϕ, R
A
(˜ z)ψ` = 'ϕ, R
A
(z)R
A
(˜ z)ψ`
=
1
z − ˜ z
('ϕ, R
A
(z)ψ` −'ϕ, R
A
(˜ z)ψ`)
=
1
z − ˜ z
R
1
λ −z
−
1
λ − ˜ z
dµ
ϕ,ψ
(λ) =
R
1
λ − ˜ z
dµ
ϕ,ψ
(λ)
λ −z
(3.46)
implying dµ
R
A
(z
∗
)ϕ,ψ
(λ) = (λ − z)
−1
dµ
ϕ,ψ
(λ) since a Herglotz function is
uniquely determined by its measure. Secondly we compute
R
1
λ −z
dµ
ϕ,P
A
(Ω)ψ
(λ) = 'ϕ, R
A
(z)P
A
(Ω)ψ` = 'R
A
(z
∗
)ϕ, P
A
(Ω)ψ`
=
R
χ
Ω
(λ)dµ
R
A
(z
∗
)ϕ,ψ
(λ) =
R
1
λ −z
χ
Ω
(λ)dµ
ϕ,ψ
(λ)
implying dµ
ϕ,P
A
(Ω)ψ
(λ) = χ
Ω
(λ)dµ
ϕ,ψ
(λ). Equivalently we have
'ϕ, P
A
(Ω
1
)P
A
(Ω
2
)ψ` = 'ϕ, P
A
(Ω
1
∩ Ω
2
)ψ` (3.47)
since χ
Ω
1
χ
Ω
2
= χ
Ω
1
∩Ω
2
. In particular, choosing Ω
1
= Ω
2
, we see that
P
A
(Ω
1
) is a projector.
The relation P
A
(R) = I follows from (3.93) below and Lemma 2.18 which
imply µ
ψ
(R) = ψ
2
.
Now let Ω =
¸
∞
n=1
Ω
n
with Ω
n
∩ Ω
m
= ∅ for n = m. Then
n
¸
j=1
'ψ, P
A
(Ω
j
)ψ` =
n
¸
j=1
µ
ψ
(Ω
j
) → 'ψ, P
A
(Ω)ψ` = µ
ψ
(Ω) (3.48)
by σadditivity of µ
ψ
. Hence P
A
is weakly σadditive which implies strong
σadditivity, as pointed out earlier.
Now we can prove the spectral theorem for selfadjoint operators.
84 3. The spectral theorem
Theorem 3.6 (Spectral theorem). To every selfadjoint operator A there
corresponds a unique projection valued measure P
A
such that
A =
R
λdP
A
(λ). (3.49)
Proof. Existence has already been established. Moreover, Lemma 3.4 shows
that P
A
((λ−z)
−1
) = R
A
(z), z ∈ C`R. Since the measures µ
ϕ,ψ
are uniquely
determined by the resolvent and the projection valued measure is uniquely
determined by the measures µ
ϕ,ψ
we are done.
The quadratic form of A is given by
q
A
(ψ) =
R
λdµ
ψ
(λ) (3.50)
and can be deﬁned for every ψ in the form domain selfadjoint operator
Q(A) = ¦ψ ∈ H[
R
[λ[dµ
ψ
(λ) < ∞¦ (3.51)
(which is larger than the domain D(A) = ¦ψ ∈ H[
R
λ
2
dµ
ψ
(λ) < ∞¦). This
extends our previous deﬁnition for nonnegative operators.
Note, that if Aand
˜
Aare unitarily equivalent as in (3.30), then UR
A
(z) =
R
˜
A
(z)U and hence
dµ
ψ
= d˜ µ
Uψ
. (3.52)
In particular, we have UP
A
(f) = P
˜
A
(f)U, UD(P
A
(f)) = D(P
˜
A
(f)).
Finally, let us give a characterization of the spectrum of A in terms of
the associated projectors.
Theorem 3.7. The spectrum of A is given by
σ(A) = ¦λ ∈ R[P
A
((λ −ε, λ +ε)) = 0 for all ε > 0¦. (3.53)
Proof. Let Ω
n
= (λ
0
−
1
n
, λ
0
+
1
n
). Suppose P
A
(Ω
n
) = 0. Then we can ﬁnd
a ψ
n
∈ P
A
(Ω
n
)H with ψ
n
 = 1. Since
(A−λ
0
)ψ
n

2
= (A−λ
0
)P
A
(Ω
n
)ψ
n

2
=
R
(λ −λ
0
)
2
χ
Ωn
(λ)dµ
ψn
(λ) ≤
1
n
2
(3.54)
we conclude λ
0
∈ σ(A) by Lemma 2.12.
Conversely, if P
A
((λ
0
−ε, λ
0
+ε)) = 0, set f
ε
(λ) = χ
R\(λ
0
−ε,λ
0
+ε)
(λ)(λ−
λ
0
)
−1
. Then
(A−λ
0
)P
A
(f
ε
) = P
A
((λ −λ
0
)f
ε
(λ)) = P
A
(R`(λ
0
−ε, λ
0
+ε)) = I. (3.55)
Similarly P
A
(f
ε
)(A−λ
0
) = I[
D(A)
and hence λ
0
∈ ρ(A).
3.2. More on Borel measures 85
Thus P
A
((λ
1
, λ
2
)) = 0 if and only if (λ
1
, λ
2
) ⊆ ρ(A) and we have
P
A
(σ(A)) = I and P
A
(R ∩ ρ(A)) = 0 (3.56)
and consequently
P
A
(f) = P
A
(σ(A))P
A
(f) = P
A
(χ
σ(A)
f). (3.57)
In other words, P
A
(f) is not aﬀected by the values of f on R`σ(A)!
It is clearly more intuitive to write P
A
(f) = f(A) and we will do so from
now on. This notation is justiﬁed by the elementary observation
P
A
(
n
¸
j=0
α
j
λ
j
) =
n
¸
j=0
α
j
A
j
. (3.58)
Moreover, this also shows that if A is bounded and f(A) can be deﬁned via
a convergent power series, then this agrees with our present deﬁnition by
Theorem 3.1.
Problem 3.1. Show that a selfadjoint operator P is a projection if and
only if σ(P) ⊆ ¦0, 1¦.
Problem 3.2. Show that (3.7) is a projection valued measure. What is the
corresponding operator?
Problem 3.3. Show that P(λ) satisﬁes the properties (i)–(iv).
Problem 3.4 (Polar decomposition). Let A be a bounded operator and set
[A[ =
√
A
∗
A. Show that
[A[ψ = Aψ.
Conclude that Ker(A) = Ker([A[) = Ran([A[)
⊥
and that
U =
ϕ = [A[ψ → Aψ if ϕ ∈ Ran([A[)
ϕ → 0 if ϕ ∈ Ker([A[)
extends to a welldeﬁned partial isometry, that is, U : Ker(U)
⊥
→ H is
norm preserving.
In particular, we have the polar decomposition
A = U[A[.
Problem 3.5. Compute [A[ =
√
A
∗
A of A = 'ϕ, .`ψ.
3.2. More on Borel measures
Section 3.1 showed that in order to understand selfadjoint operators, one
needs to understand multiplication operators on L
2
(R, dµ), where dµ is a
ﬁnite Borel measure. This is the purpose of the present section.
86 3. The spectral theorem
The set of all growth points, that is,
σ(µ) = ¦λ ∈ R[µ((λ −ε, λ +ε)) > 0 for all ε > 0¦, (3.59)
is called the spectrum of µ. Invoking Morea’s together with Fubini’s theorem
shows that the Borel transform
F(z) =
R
1
λ −z
dµ(λ) (3.60)
is holomorphic for z ∈ C`σ(µ). The converse following from Stieltjes inver
sion formula. Associated with this measure is the operator
Af(λ) = λf(λ), D(A) = ¦f ∈ L
2
(R, dµ)[λf(λ) ∈ L
2
(R, dµ)¦. (3.61)
By Theorem 3.7 the spectrum of A is precisely the spectrum of µ, that is,
σ(A) = σ(µ). (3.62)
Note that 1 ∈ L
2
(R, dµ) is a cyclic vector for A and that
dµ
g,f
(λ) = g(λ)
∗
f(λ)dµ(λ). (3.63)
Now what can we say about the function f(A) (which is precisely the
multiplication operator by f) of A? We are only interested in the case where
f is realvalued. Introduce the measure
(f
µ)(Ω) = µ(f
−1
(Ω)), (3.64)
then
R
g(λ)d(f
µ)(λ) =
R
g(f(λ))dµ(λ). (3.65)
In fact, it suﬃces to check this formula for simple functions g which follows
since χ
Ω
◦ f = χ
f
−1
(Ω)
. In particular, we have
P
f(A)
(Ω) = χ
f
−1
(Ω)
. (3.66)
It is tempting to conjecture that f(A) is unitarily equivalent to multi
plication by λ in L
2
(R, d(f
µ)) via the map
L
2
(R, d(f
µ)) → L
2
(R, dµ), g → g ◦ f. (3.67)
However, this map is only unitary if its range is L
2
(R, dµ), that is, if f is
injective.
Lemma 3.8. Suppose f is injective, then
U : L
2
(R, dµ) → L
2
(R, d(f
µ)), g → g ◦ f
−1
(3.68)
is a unitary map such that Uf(λ) = λ.
3.2. More on Borel measures 87
Example. Let f(λ) = λ
2
, then (g ◦ f)(λ) = g(λ
2
) and the range of the
above map is given by the symmetric functions. Note that we can still
get a unitary map L
2
(R, d(f
µ)) ⊕ L
2
(R, χd(f
µ)) → L
2
(R, dµ), (g
1
, g
2
) →
g
1
(λ
2
) +g
2
(λ
2
)(χ(λ) −χ(−λ)), where χ = χ
(0,∞)
.
Lemma 3.9. Let f be realvalued. The spectrum of f(A) is given by
σ(f(A)) = σ(f
µ). (3.69)
In particular,
σ(f(A)) ⊆ f(σ(A)), (3.70)
where equality holds if f is continuous and the closure can be dropped if, in
addition, σ(A) is bounded (i.e., compact).
Proof. If λ
0
∈ σ(f
µ), then the sequence g
n
= µ(Ω
n
)
−1/2
χ
Ωn
, Ω
n
=
f
−1
((λ
0
−
1
n
, λ
0
+
1
n
)), satisﬁes g
n
 = 1, (f(A) −λ
0
)g
n
 < n
−1
and hence
λ
0
∈ σ(f(A)). Conversely, if λ
0
∈ σ(f
µ), then µ(Ω
n
) = 0 for some n and
hence we can change f on Ω
n
such that f(R) ∩(λ
0
−
1
n
, λ
0
+
1
n
) = ∅ without
changing the corresponding operator. Thus (f(A) −λ
0
)
−1
= (f(λ) −λ
0
)
−1
exists and is bounded, implying λ
0
∈ σ(f(A)).
If f is continuous, f
−1
(f(λ) − ε, f(λ) + ε) contains an open interval
around λ and hence f(λ) ∈ σ(f(A)) if λ ∈ σ(A). If, in addition, σ(A) is
compact, then f(σ(A)) is compact and hence closed.
If two operators with simple spectrum are unitarily equivalent can be
read oﬀ from the corresponding measures:
Lemma 3.10. Let A
1
, A
2
be selfadjoint operators with simple spectrum and
corresponding spectral measures µ
1
and µ
2
of cyclic vectors. Then A
1
and
A
2
are unitarily equivalent if and only if µ
1
and µ
2
are mutually absolutely
continuous.
Proof. Without restriction we can assume that A
j
is multiplication by λ
in L
2
(R, dµ
j
). Let U : L
2
(R, dµ
1
) → L
2
(R, dµ
2
) be a unitary map such that
UA
1
= A
2
U. Then we also have Uf(A
1
) = f(A
2
)U for any bounded Borel
Function and hence
Uf(λ) = Uf(λ) 1 = f(λ)U(1)(λ) (3.71)
and thus U is multiplication by u(λ) = U(1)(λ). Moreover, since U is
unitary we have
µ
1
(Ω) =
R
[χ
Ω
[
2
dµ
1
=
R
[uχ
Ω
[
2
dµ
2
=
Ω
[u[
2
dµ
2
, (3.72)
that is, dµ
1
= [u[
2
dµ
2
. Reversing the role of A
1
and A
2
we obtain dµ
2
=
[v[
2
dµ
1
, where v = U
−1
1.
88 3. The spectral theorem
The converse is left as an exercise (Problem 3.10.)
Next we recall the unique decomposition of µ with respect to Lebesgue
measure,
dµ = dµ
ac
+dµ
s
, (3.73)
where µ
ac
is absolutely continuous with respect to Lebesgue measure (i.e.,
we have µ
ac
(B) = 0 for all B with Lebesgue measure zero) and µ
s
is singular
with respect to Lebesgue measure (i.e., µ
s
is supported, µ
s
(R`B) = 0, on
a set B with Lebesgue measure zero). The singular part µ
s
can be further
decomposed into a (singularly) continuous and a pure point part,
dµ
s
= dµ
sc
+dµ
pp
, (3.74)
where µ
sc
is continuous on R and µ
pp
is a step function. Since the measures
dµ
ac
, dµ
sc
, and dµ
pp
are mutually singular, they have mutually disjoint
supports M
ac
, M
sc
, and M
pp
. Note that these sets are not unique. We will
choose them such that M
pp
is the set of all jumps of µ(λ) and such that M
sc
has Lebesgue measure zero.
To the sets M
ac
, M
sc
, and M
pp
correspond projectors P
ac
= χ
Mac
(A),
P
sc
= χ
Msc
(A), and P
pp
= χ
Mpp
(A) satisfying P
ac
+ P
sc
+ P
pp
= I. In
other words, we have a corresponding direct sum decomposition of both our
Hilbert space
L
2
(R, dµ) = L
2
(R, dµ
ac
) ⊕L
2
(R, dµ
sc
) ⊕L
2
(R, dµ
pp
) (3.75)
and our operator A
A = (AP
ac
) ⊕(AP
sc
) ⊕(AP
pp
). (3.76)
The corresponding spectra, σ
ac
(A) = σ(µ
ac
), σ
sc
(A) = σ(µ
sc
), and σ
pp
(A) =
σ(µ
pp
) are called the absolutely continuous, singularly continuous, and pure
point spectrum of A, respectively.
It is important to observe that σ
pp
(A) is in general not equal to the set
of eigenvalues
σ
p
(A) = ¦λ ∈ R[λ is an eigenvalue of A¦ (3.77)
since we only have σ
pp
(A) = σ
p
(A).
Example. let H =
2
(N) and let A be given by Aδ
n
=
1
n
δ
n
, where δ
n
is the sequence which is 1 at the n’th place and zero else (that is, A is
a diagonal matrix with diagonal elements
1
n
). Then σ
p
(A) = ¦
1
n
[n ∈ N¦
but σ(A) = σ
pp
(A) = σ
p
(A) ∪ ¦0¦. To see this, just observe that δ
n
is the
eigenvector corresponding to the eigenvalue
1
n
and for z ∈ σ(A) we have
R
A
(z)δ
n
=
n
1−nz
δ
n
. At z = 0 this formula still gives the inverse of A, but
it is unbounded and hence 0 ∈ σ(A) but 0 ∈ σ
p
(A). Since a continuous
measure cannot live on a single point and hence also not on a countable set,
we have σ
ac
(A) = σ
sc
(A) = ∅.
3.3. Spectral types 89
Example. An example with purely absolutely continuous spectrum is given
by taking µ to be Lebesgue measure. An example with purely singularly
continuous spectrum is given by taking µ to be the Cantor measure.
Problem 3.6. Construct a multiplication operator on L
2
(R) which has
dense point spectrum.
Problem 3.7. Let λ be Lebesgue measure on R. Show that if f ∈ AC(R)
with f
> 0, then
d(f
λ) =
1
f
(λ)
dλ. (3.78)
Problem 3.8. Let dµ(λ) = χ
[0,1]
(λ)dλ and f(λ) = χ
(−∞,t]
, t ∈ R. Compute
f
µ.
Problem 3.9. Let A be the multiplication operator by the Cantor function
in L
2
(0, 1). Compute the spectrum of A. Determine the spectral types.
Problem 3.10. Show the missing direction in the proof of Lemma 3.10.
3.3. Spectral types
Our next aim is to transfer the results of the previous section to arbitrary
selfadjoint operators A using Lemma 3.3. To do this we will need a spectral
measure which contains the information from all measures in a spectral basis.
This will be the case if there is a vector ψ such that for every ϕ ∈ H its
spectral measureµ
ϕ
is absolutely continuous with respect to µ
ψ
. Such a
vector will be called a maximal spectral vector of A and µ
ψ
will be
called a maximal spectral measure of A.
Lemma 3.11. For every selfadjoint operator A there is a maximal spectral
vector.
Proof. Let ¦ψ
j
¦
j∈J
be a spectral basis and choose nonzero numbers ε
j
with
¸
j∈J
[ε
j
[
2
= 1. Then I claim that
ψ =
¸
j∈J
ε
j
ψ
j
(3.79)
is a maximal spectral vector. Let ϕ be given, then we can write it as ϕ =
¸
j
f
j
(A)ψ
j
and hence dµ
ϕ
=
¸
j
[f
j
[
2
dµ
ψ
j
. But µ
ψ
(Ω) =
¸
j
[ε
j
[
2
µ
ψ
j
(Ω) =
0 implies µ
ψ
j
(Ω) = 0 for every j ∈ J and thus µ
ϕ
(Ω) = 0.
A set ¦ψ
j
¦ of spectral vectors is called ordered if ψ
k
is a maximal
spectral vector for A restricted to (
¸
k−1
j=1
H
ψ
j
)
⊥
. As in the unordered case
one can show
90 3. The spectral theorem
Theorem 3.12. For every selfadjoint operator there is an ordered spectral
basis.
Observe that if ¦ψ
j
¦ is an ordered spectral basis, then µ
ψ
j+1
is absolutely
continuous with respect to µ
ψ
j
.
If µ is a maximal spectral measure we have σ(A) = σ(µ) and the follow
ing generalization of Lemma 3.9 holds.
Theorem 3.13 (Spectral mapping). Let µ be a maximal spectral measure
and let f be realvalued. Then the spectrum of f(A) is given by
σ(f(A)) = ¦λ ∈ R[µ(f
−1
(λ −ε, λ +ε)) > 0 for all ε > 0¦. (3.80)
In particular,
σ(f(A)) ⊆ f(σ(A)), (3.81)
where equality holds if f is continuous and the closure can be dropped if, in
addition, σ(A) is bounded.
Next, we want to introduce the splitting (3.75) for arbitrary selfadjoint
operators A. It is tempting to pick a spectral basis and treat each summand
in the direct sum separately. However, since it is not clear that this approach
is independent of the spectral basis chosen, we use the more sophisticated
deﬁnition
H
ac
= ¦ψ ∈ H[µ
ψ
is absolutely continuous¦,
H
sc
= ¦ψ ∈ H[µ
ψ
is singularly continuous¦,
H
pp
= ¦ψ ∈ H[µ
ψ
is pure point¦. (3.82)
Lemma 3.14. We have
H = H
ac
⊕H
sc
⊕H
pp
. (3.83)
There are Borel sets M
xx
such that the projector onto H
xx
is given by P
xx
=
χ
Mxx
(A), xx ∈ ¦ac, sc, pp¦. In particular, the subspaces H
xx
reduce A. For
the sets M
xx
one can choose the corresponding supports of some maximal
spectral measure µ.
Proof. We will use the unitary operator U of Lemma 3.3. Pick ϕ ∈ H and
write ϕ =
¸
n
ϕ
n
with ϕ
n
∈ H
ψn
. Let f
n
= Uϕ
n
, then, by construction
of the unitary operator U, ϕ
n
= f
n
(A)ψ
n
and hence dµ
ϕn
= [f
n
[
2
dµ
ψn
.
Moreover, since the subspaces H
ψn
are orthogonal, we have
dµ
ϕ
=
¸
n
[f
n
[
2
dµ
ψn
(3.84)
and hence
dµ
ϕ,xx
=
¸
n
[f
n
[
2
dµ
ψn,xx
, xx ∈ ¦ac, sc, pp¦. (3.85)
3.4. Appendix: The Herglotz theorem 91
This shows
UH
xx
=
n
L
2
(R, dµ
ψn,xx
), xx ∈ ¦ac, sc, pp¦ (3.86)
and reduces our problem to the considerations of the previous section.
The absolutely continuous, singularly continuous, and pure point
spectrum of A are deﬁned as
σ
ac
(A) = σ(A[
Hac
), σ
sc
(A) = σ(A[
Hsc
), and σ
pp
(A) = σ(A[
Hpp
),
(3.87)
respectively. If µ is a maximal spectral measure we have σ
ac
(A) = σ(µ
ac
),
σ
sc
(A) = σ(µ
sc
), and σ
pp
(A) = σ(µ
pp
).
If A and
˜
A are unitarily equivalent via U, then so are A[
Hxx
and
˜
A[
˜
Hxx
by (3.52). In particular, σ
xx
(A) = σ
xx
(
˜
A).
Problem 3.11. Compute σ(A), σ
ac
(A), σ
sc
(A), and σ
pp
(A) for the multi
plication operator A =
1
1+x
2
in L
2
(R). What is its spectral multiplicity?
3.4. Appendix: The Herglotz theorem
A holomorphic function F : C
+
→C
+
, C
±
= ¦z ∈ C[ ±Im(z) > 0¦, is called
a Herglotz function. We can deﬁne F on C
−
using F(z
∗
) = F(z)
∗
.
Suppose µ is a ﬁnite Borel measure. Then its Borel transform is deﬁned
via
F(z) =
R
dµ(λ)
λ −z
. (3.88)
Theorem 3.15. The Borel transform of a ﬁnite Borel measure is a Herglotz
function satisfying
[F(z)[ ≤
µ(R)
Im(z)
, z ∈ C
+
. (3.89)
Moreover, the measure µ can be reconstructed via Stieltjes inversion formula
1
2
(µ((λ
1
, λ
2
)) +µ([λ
1
, λ
2
])) = lim
ε↓0
1
π
λ
2
λ
1
Im(F(λ + iε))dλ. (3.90)
Proof. By Morea’s and Fubini’s theorem, F is holomorphic on C
+
and the
remaining properties follow from 0 < Im((λ−z)
−1
) and [λ−z[
−1
≤ Im(z)
−1
.
Stieltjes inversion formula follows from Fubini’s theorem and the dominated
convergence theorem since
1
2πi
λ
2
λ
1
(
1
x −λ −iε
−
1
x −λ + iε
)dλ →
1
2
χ
[λ
1
,λ
2
]
(x) +χ
(λ
1
,λ
2
)
(x)
(3.91)
pointwise.
92 3. The spectral theorem
Observe
Im(F(z)) = Im(z)
R
dµ(λ)
[λ −z[
2
(3.92)
and
lim
λ→∞
λIm(F(iλ)) = µ(R). (3.93)
The converse of the previous theorem is also true
Theorem 3.16. Suppose F is a Herglotz function satisfying
[F(z)[ ≤
M
Im(z)
, z ∈ C
+
. (3.94)
Then there is a unique Borel measure µ, satisfying µ(R) ≤ M, such that F
is the Borel transform of µ.
Proof. We abbreviate F(z) = v(z) +i w(z) and z = x+i y. Next we choose
a contour
Γ = ¦x + iε +λ[λ ∈ [−R, R]¦ ∪ ¦x + iε +Re
iϕ
[ϕ ∈ [0, π]¦. (3.95)
and note that z lies inside Γ and z
∗
+ 2iε lies outside Γ if 0 < ε < y < R.
Hence we have by Cauchy’s formula
F(z) =
1
2πi
Γ
1
ζ −z
−
1
ζ −z
∗
−2iε
F(ζ)dζ. (3.96)
Inserting the explicit form of Γ we see
F(z) =
1
π
R
−R
y −ε
λ
2
+ (y −ε)
2
F(x + iε +λ)dλ
+
i
π
π
0
y −ε
R
2
e
2iϕ
+ (y −ε)
2
F(x + iε +Re
iϕ
)Re
iϕ
dϕ. (3.97)
The integral over the semi circle vanishes as R → ∞ and hence we obtain
F(z) =
1
π
R
y −ε
(λ −x)
2
+ (y −ε)
2
F(λ + iε)dλ (3.98)
and taking imaginary parts
w(z) =
R
φ
ε
(λ)w
ε
(λ)dλ, (3.99)
where φ
ε
(λ) = (y−ε)/((λ−x)
2
+(y−ε)
2
) and w
ε
(λ) = w(λ+iε)/π. Letting
y → ∞ we infer from our bound
R
w
ε
(λ)dλ ≤ M. (3.100)
In particular, since [φ
ε
(λ) −φ
0
(λ)[ ≤ const ε we have
w(z) = lim
ε↓0
R
φ
0
(λ)dµ
ε
(λ), (3.101)
3.4. Appendix: The Herglotz theorem 93
where µ
ε
(λ) =
λ
−∞
w
ε
(x)dx. It remains to establish that the monotone
functions µ
ε
converge properly. Since 0 ≤ µ
ε
(λ) ≤ M, there is a convergent
subsequence for ﬁxed λ. Moreover, by the standard diagonal trick, there
is even a subsequence ε
n
such that µ
εn
(λ) converges for each rational λ.
For irrational λ we set µ(λ
0
) = inf
λ≥λ
0
¦µ(λ)[λ rational¦. Then µ(λ) is
monotone, 0 ≤ µ(λ
1
) ≤ µ(λ
2
) ≤ M, λ
1
≤ λ
2
, and we claim
w(z) =
R
φ
0
(λ)dµ(λ). (3.102)
Fix δ > 0 and let λ
1
< λ
2
< < λ
m+1
be rational numbers such that
[λ
j+1
−λ
j
[ ≤ δ and λ
1
≤ x −
y
3
δ
, λ
m+1
≥ x +
y
3
δ
. (3.103)
Then (since [φ
0
(λ)[ ≤ y
−2
)
[φ
0
(λ) −φ
0
(λ
j
)[ ≤
δ
y
2
, λ
j
≤ λ ≤ λ
j+1
, (3.104)
and
[φ
0
(λ)[ ≤
δ
y
2
, λ ≤ λ
1
or λ
m+1
≤ λ. (3.105)
Now observe
[
R
φ
0
(λ)dµ(λ) −
R
φ
0
(λ)dµ
εn
(λ)[ ≤
[
R
φ
0
(λ)dµ(λ) −
m
¸
j=1
φ
0
(λ
j
)(µ(λ
j+1
) −µ(λ
j
))[
+[
m
¸
j=1
φ
0
(λ
j
)(µ(λ
j+1
) −µ(λ
j
) −µ
εn
(λ
j+1
) +µ
εn
(λ
j
))
+[
R
φ
0
(λ)dµ
εn
(λ) −
m
¸
j=1
φ
0
(λ
j
)(µ
εn
(λ
j+1
) −µ
εn
(λ
j
))[ (3.106)
The ﬁrst and third term can be bounded by 2Mδ/y
2
. Moreover, since
φ
0
(y) ≤ 1/y we can ﬁnd an N ∈ N such that
[µ(λ
j
) −µ
εn
(λ
j
)[ ≤
y
2m
δ, n ≥ N, (3.107)
and hence the second term is bounded by δ. In summary, the diﬀerence in
(3.106) can be made arbitrarily small.
Now F(z) and
R
(λ−z)
−1
dµ(λ) have the same imaginary part and thus
they only diﬀer by a real constant. By our bound this constant must be
zero.
The RadonNikodym derivative of µ can be obtained from the boundary
values of F.
94 3. The spectral theorem
Theorem 3.17. Let µ be a ﬁnite Borel measure and F its Borel transform,
then
(Dµ)(λ) ≤ liminf
ε↓0
1
π
F(λ + iε) ≤ limsup
ε↓0
1
π
F(λ + iε) ≤ (Dµ)(λ). (3.108)
Proof. We need to estimate
Im(F(λ + iε)) =
R
K
ε
(t)dµ(t), K
ε
(t) =
ε
t
2
+ε
2
. (3.109)
We ﬁrst split the integral into two parts
Im(F(λ+iε)) =
I
δ
K
ε
(t−λ)dµ(t)+
R\I
δ
K
ε
(t−λ)µ(t), I
δ
= (λ−δ, λ+δ).
(3.110)
Clearly the second part can be estimated by
R\I
δ
K
ε
(t −λ)µ(t) ≤ K
ε
(δ)µ(R). (3.111)
To estimate the ﬁrst part we integrate
K
ε
(s) ds dµ(t) (3.112)
over the triangle ¦(s, t)[λ − s < t < λ + s, 0 < s < δ¦ = ¦(s, t)[λ − δ < t <
λ +δ, t −λ < s < δ¦ and obtain
δ
0
µ(I
s
)K
ε
(s)ds =
I
δ
(K(δ) −K
ε
(t −λ))dµ(t). (3.113)
Now suppose there is are constants c and C such that c ≤
µ(Is)
2s
≤ C,
0 ≤ s ≤ δ, then
2c arctan(
δ
ε
) ≤
I
δ
K
ε
(t −λ)dµ(t) ≤ 2C arctan(
δ
ε
) (3.114)
since
δK
ε
(δ) +
δ
0
−sK
ε
(s)ds = arctan(
δ
ε
). (3.115)
Thus the claim follows combining both estimates.
As a consequence of Theorem A.34 and Theorem A.35 we obtain
Theorem 3.18. Let µ be a ﬁnite Borel measure and F its Borel transform,
then the limit
Im(F(λ)) = lim
ε↓0
1
π
Im(F(λ + iε)) (3.116)
exists a.e. with respect to both µ and Lebesgue measure (ﬁnite or inﬁnite)
and
(Dµ)(λ) =
1
π
Im(F(λ)) (3.117)
whenever (Dµ)(λ) exists.
3.4. Appendix: The Herglotz theorem 95
Moreover, the set ¦λ[F(λ) = ∞¦ is a support for the singularly and
¦λ[F(λ) < ∞¦ is a support for the absolutely continuous part.
In particular,
Corollary 3.19. The measure µ is purely absolutely continuous on I if
limsup
ε↓0
Im(F(λ + iε)) < ∞ for all λ ∈ I.
Chapter 4
Applications of the
spectral theorem
This chapter can be mostly skipped on ﬁrst reading. You might want to have a
look at the ﬁrst section and the come back to the remaining ones later.
Now let us show how the spectral theorem can be used. We will give a
few typical applications:
Firstly we will derive an operator valued version of of Stieltjes’ inversion
formula. To do this, we need to show how to integrate a family of functions
of A with respect to a parameter. Moreover, we will show that these integrals
can be evaluated by computing the corresponding integrals of the complex
valued functions.
Secondly we will consider commuting operators and show how certain
facts, which are known to hold for the resolvent of an operator A, can be
established for a larger class of functions.
Then we will show how the eigenvalues below the essential spectrum and
dimension of RanP
A
(Ω) can be estimated using the quadratic form.
Finally, we will investigate tensor products of operators.
4.1. Integral formulas
We begin with the ﬁrst task by having a closer look at the projector P
A
(Ω).
They project onto subspaces corresponding to expectation values in the set
Ω. In particular, the number
'ψ, χ
Ω
(A)ψ` (4.1)
97
98 4. Applications of the spectral theorem
is the probability for a measurement of a to lie in Ω. In addition, we have
'ψ, Aψ` =
Ω
λdµ
ψ
(λ) ∈ hull(Ω), ψ ∈ P
A
(Ω)H, ψ = 1, (4.2)
where hull(Ω) is the convex hull of Ω.
The space Ranχ
{λ
0
}
(A) is called the eigenspace corresponding to λ
0
since we have
'ϕ, Aψ` =
R
λχ
{λ
0
}
(λ)dµ
ϕ,ψ
(λ) = λ
0
R
dµ
ϕ,ψ
(λ) = λ
0
'ϕ, ψ` (4.3)
and hence Aψ = λ
0
ψ for all ψ ∈ Ranχ
{λ
0
}
(A). The dimension of the
eigenspace is called the multiplicity of the eigenvalue.
Moreover, since
lim
ε↓0
−iε
λ −λ
0
−iε
= χ
{λ
0
}
(λ) (4.4)
we infer from Theorem 3.1 that
lim
ε↓0
−iεR
A
(λ
0
+ iε)ψ = χ
{λ
0
}
(A)ψ. (4.5)
Similarly, we can obtain an operator valued version of Stieltjes’ inversion
formula. But ﬁrst we need to recall a few facts from integration in Banach
spaces.
We will consider the case of mappings f : I → X where I = [t
0
, t
1
] ⊂ R is
a compact interval and X is a Banach space. As before, a function f : I → X
is called simple if the image of f is ﬁnite, f(I) = ¦x
i
¦
n
i=1
, and if each inverse
image f
−1
(x
i
), 1 ≤ i ≤ n, is a Borel set. The set of simple functions S(I, X)
forms a linear space and can be equipped with the sup norm
f
∞
= sup
t∈I
f(t). (4.6)
The corresponding Banach space obtained after completion is called the set
of regulated functions R(I, X).
Observe that C(I, X) ⊂ R(I, X). In fact, consider the simple function
f
n
=
¸
n−1
i=0
f(s
i
)χ
[s
i
,s
i+1
)
, where s
i
= t
0
+ i
t
1
−t
0
n
. Since f ∈ C(I, X) is
uniformly continuous, we infer that f
n
converges uniformly to f.
For f ∈ S(I, X) we can deﬁne a linear map
: S(I, X) → X by
I
f(t)dt =
n
¸
i=1
x
i
[f
−1
(x
i
)[, (4.7)
where [Ω[ denotes the Lebesgue measure of Ω. This map satisﬁes

I
f(t)dt ≤ f
∞
(t
1
−t
0
) (4.8)
4.1. Integral formulas 99
and hence it can be extended uniquely to a linear map
: R(I, X) → X
with the same norm (t
1
−t
0
) by Theorem 0.24. We even have

I
f(t)dt ≤
I
f(t)dt, (4.9)
which clearly holds for f ∈ S(I, X) und thus for all f ∈ R(I, X) by conti
nuity. In addition, if ∈ X
∗
is a continuous linear functional, then
(
I
f(t)dt) =
I
(f(t))dt, f ∈ R(I, X). (4.10)
In particular, if A(t) ∈ R(I, L(H)), then
I
A(t)dt
ψ =
I
(A(t)ψ)dt. (4.11)
If I = R, we say that f : I → X is integrable if f ∈ R([−r, r], X) for all
r > 0 and if f(t) is integrable. In this case we can set
R
f(t)dt = lim
r→∞
[−r,r]
f(t)dt (4.12)
and (4.9) and (4.10) still hold.
We will use the standard notation
t
3
t
2
f(s)ds =
I
χ
(t
2
,t
3
)
(s)f(s)ds and
t
2
t
3
f(s)ds = −
t
3
t
2
f(s)ds.
We write f ∈ C
1
(I, X) if
d
dt
f(t) = lim
ε→0
f(t +ε) −f(t)
ε
(4.13)
exists for all t ∈ I. In particular, if f ∈ C(I, X), then F(t) =
t
t
0
f(s)ds ∈
C
1
(I, X) and dF/dt = f as can be seen from
F(t +ε) −F(t) −f(t)ε = 
t+ε
t
(f(s) −f(t))ds ≤ [ε[ sup
s∈[t,t+ε]
f(s) −f(t).
(4.14)
The important facts for us are the following two results.
Lemma 4.1. Suppose f : I R →C is a bounded Borel function such that
f(., λ) and set F(λ) =
I
f(t, λ)dt. Let A be selfadjoint. Then f(t, A) ∈
R(I, L(H)) and
F(A) =
I
f(t, A)dt respectively F(A)ψ =
I
f(t, A)ψ dt. (4.15)
Proof. That f(t, A) ∈ R(I, L(H)) follows from the spectral theorem, since
it is no restriction to assume that A is multiplication by λ in some L
2
space.
100 4. Applications of the spectral theorem
We compute
'ϕ, (
I
f(t, A)dt)ψ` =
I
'ϕ, f(t, A)ψ`dt
=
I
R
f(t, λ)dµ
ϕ,ψ
(λ)dt
=
R
I
f(t, λ)dt dµ
ϕ,ψ
(λ)
=
R
F(λ)dµ
ϕ,ψ
(λ) = 'ϕ, F(A)ψ` (4.16)
by Fubini’s theorem and hence the ﬁrst claim follows.
Lemma 4.2. Suppose f : R →L(H) is integrable and A ∈ L(H). Then
A
R
f(t)dt =
R
Af(t)dt respectively
R
f(t)dtA =
R
f(t)Adt. (4.17)
Proof. It suﬃces to prove the case where f is simple and of compact sup
port. But for such functions the claim is straightforward.
Now we can prove Stone’s formula.
Theorem 4.3 (Stone’s formula). Let A be selfadjoint, then
1
2πi
λ
2
λ
1
(R
A
(λ + iε) −R
A
(λ −iε))dλ →
1
2
(P
A
([λ
1
, λ
2
]) +P
A
((λ
1
, λ
2
)))
(4.18)
strongly.
Proof. The result follows combining Lemma 4.1 with Theorem 3.1 and
(3.91).
Problem 4.1. Let Γ be a diﬀerentiable Jordan curve in ρ(A). Show
χ
Ω
(A) =
Γ
R
A
(z)dz, (4.19)
where Ω is the intersection of the interior of Γ with R.
4.2. Commuting operators
Now we come to commuting operators. As a preparation we can now prove
Lemma 4.4. Let K ⊆ R be closed. And let C
∞
(K) be the set of all contin
uous functions on K which vanish at ∞ (if K is unbounded) with the sup
norm. The ∗algebra generated by the function
λ →
1
λ −z
(4.20)
for one z ∈ C`K is dense in C
∞
(K).
4.2. Commuting operators 101
Proof. If K is compact, the claim follows directly from the complex Stone–
Weierstraß theorem since (λ
1
−z)
−1
= (λ
2
−z)
−1
implies λ
1
= λ
2
. Otherwise,
replace K by
˜
K = K∪¦∞¦, which is compact, and set (∞−z)
−1
= 0. Then
we can again apply the complex Stone–Weierstraß theorem to conclude that
our ∗subalgebra is equal to ¦f ∈ C(
˜
K)[f(∞) = 0¦ which is equivalent to
C
∞
(K).
We say that two bounded operators A, B commute if
[A, B] = AB −BA = 0. (4.21)
If A or B is unbounded, we soon run into trouble with this deﬁnition since
the above expression might not even make sense for any nonzero vector (e.g.,
take B = 'ϕ, .`ψ with ψ ∈ D(A)). To avoid this nuisance we will replace A
by a bounded function of A. A good candidate is the resolvent. Hence if A
is selfadjoint and B is bounded we will say that A and B commute if
[R
A
(z), B] = [R
A
(z
∗
), B] = 0 (4.22)
for one z ∈ ρ(A).
Lemma 4.5. Suppose A is selfadjoint and commutes with the bounded
operator B. Then
[f(A), B] = 0 (4.23)
for any bounded Borel function f. If f is unbounded, the claim holds for
any ψ ∈ D(f(A)).
Proof. Equation (4.22) tell us that (4.23) holds for any f in the ∗subalgebra
generated by R
A
(z). Since this subalgebra is dense in C
∞
(σ(A)), the claim
follows for all such f ∈ C
∞
(σ(A)). Next ﬁx ψ ∈ H and let f be bounded.
Choose a sequence f
n
∈ C
∞
(σ(A)) converging to f in L
2
(R, dµ
ψ
). Then
Bf(A)ψ = lim
n→∞
Bf
n
(A)ψ = lim
n→∞
f
n
(A)Bψ = f(A)Bψ. (4.24)
If f is unbounded, let ψ ∈ D(f(A)) and choose f
n
as in (3.24). Then
f(A)Bψ = lim
n→∞
f
n
(A)Bψ = lim
n→∞
Bf
n
(A)ψ (4.25)
shows f ∈ L
2
(R, dµ
Bψ
) (i.e., Bψ ∈ D(f(A))) and f(A)Bψ = BF(A)ψ.
Corollary 4.6. If A is selfadjoint and bounded, then (4.22) holds if and
only if (4.21) holds.
Proof. Since σ(A) is compact, we have λ ∈ C
∞
(σ(A)) and hence (4.21)
follows from (4.23) by our lemma. Conversely, since B commutes with any
polynomial of A, the claim follows from the Neumann series.
As another consequence we obtain
102 4. Applications of the spectral theorem
Theorem 4.7. Suppose A is selfadjoint and has simple spectrum. A bounded
operator B commutes with A if and only if B = f(A) for some bounded Borel
function.
Proof. Let ψ be a cyclic vector for A. By our unitary equivalence it is no
restriction to assume H = L
2
(R, dµ
ψ
). Then
Bg(λ) = Bg(λ) 1 = g(λ)(B1)(λ) (4.26)
since B commutes with the multiplication operator g(λ). Hence B is multi
plication by f(λ) = (B1)(λ).
The assumption that the spectrum of A is simple is crucial as the exam
ple A = I shows. Note also that the functions exp(−itA) can also be used
instead of resolvents.
Lemma 4.8. Suppose A is selfadjoint and B is bounded. Then B commutes
with A if and only if
[e
−iAt
, B] = 0 (4.27)
for all t ∈ R.
Proof. It suﬃces to show [
ˆ
f(A), B] = 0 for f ∈ o(R), since these functions
are dense in C
∞
(R) by the complex Stone–Weierstraß theorem. Here
ˆ
f
denotes the Fourier transform of f, see Section 7.1. But for such f we have
[
ˆ
f(A), B] =
1
√
2π
[
R
f(t)e
−iAt
dt, B] =
1
√
2π
R
f(t)[e
−iAt
, B]dt = 0 (4.28)
by Lemma 4.2.
The extension to the case where B is selfadjoint and unbounded is
straightforward. We say that A and B commute in this case if
[R
A
(z
1
), R
B
(z
2
)] = [R
A
(z
∗
1
), R
B
(z
2
)] = 0 (4.29)
for one z
1
∈ ρ(A) and one z
2
∈ ρ(B) (the claim for z
∗
2
follows by taking
adjoints). From our above analysis it follows that this is equivalent to
[e
−iAt
, e
−iBs
] = 0, t, s ∈ R, (4.30)
respectively
[f(A), g(B)] = 0 (4.31)
for arbitrary bounded Borel functions f and g.
4.3. The minmax theorem 103
4.3. The minmax theorem
In many applications a selfadjoint operator has a number of eigenvalues be
low the bottom of the essential spectrum. The essential spectrum is obtained
from the spectrum by removing all discrete eigenvalues with ﬁnite multiplic
ity (we will have a closer look at it in Section 6.2). In general there is no way
of computing the lowest eigenvalues and their corresponding eigenfunctions
explicitly. However, one often has some idea how the eigenfunctions might
approximately look like.
So suppose we have a normalized function ψ
1
which is an approximation
for the eigenfunction ϕ
1
of the lowest eigenvalue E
1
. Then by Theorem 2.15
we know that
'ψ
1
, Aψ
1
` ≥ 'ϕ
1
, Aϕ
1
` = E
1
. (4.32)
If we add some free parameters to ψ
1
, one can optimize them and obtain
quite good upper bounds for the ﬁrst eigenvalue.
But is there also something one can say about the next eigenvalues?
Suppose we know the ﬁrst eigenfunction ϕ
1
, then we can restrict A to the
orthogonal complement of ϕ
1
and proceed as before: E
2
will be the inﬁmum
over all expectations restricted to this subspace. If we restrict to the or
thogonal complement of an approximating eigenfunction ψ
1
, there will still
be a component in the direction of ϕ
1
left and hence the inﬁmum of the
expectations will be lower than E
2
. Thus the optimal choice ψ
1
= ϕ
1
will
give the maximal value E
2
.
More precisely, let ¦ϕ
j
¦
N
j=1
be an orthonormal basis for the space spanned
by the eigenfunctions corresponding to eigenvalues below the essential spec
trum. Assume they satisfy (A−E
j
)ϕ
j
= 0, where E
j
≤ E
j+1
are the eigen
values (counted according to their multiplicity). If the number of eigenvalues
N is ﬁnite we set E
j
= inf σ
ess
(A) for j > N and choose ϕ
j
orthonormal
such that (A−E
j
)ϕ
j
 ≤ ε.
Deﬁne
U(ψ
1
, . . . , ψ
n
) = ¦ψ ∈ D(A)[ ψ = 1, ψ ∈ span¦ψ
1
, . . . , ψ
n
¦
⊥
¦. (4.33)
(i) We have
inf
ψ∈U(ψ
1
,...,ψ
n−1
)
'ψ, Aψ` ≤ E
n
+O(ε). (4.34)
In fact, set ψ =
¸
n
j=1
α
j
ϕ
j
and choose α
j
such that ψ ∈ U(ψ
1
, . . . , ψ
n−1
),
then
'ψ, Aψ` =
n
¸
j=1
[α
j
[
2
E
j
+O(ε) ≤ E
n
+O(ε) (4.35)
and the claim follows.
104 4. Applications of the spectral theorem
(ii) We have
inf
ψ∈U(ϕ
1
,...,ϕ
n−1
)
'ψ, Aψ` ≥ E
n
−O(ε). (4.36)
In fact, set ψ = ϕ
n
.
Since ε can be chosen arbitrarily small we have proven
Theorem 4.9 (MinMax). Let A be selfadjoint and let E
1
≤ E
2
≤ E
3
be the eigenvalues of A below the essential spectrum respectively the inﬁmum
of the essential spectrum once there are no more eigenvalues left. Then
E
n
= sup
ψ
1
,...,ψ
n−1
inf
ψ∈U(ψ
1
,...,ψ
n−1
)
'ψ, Aψ`. (4.37)
Clearly the same result holds if D(A) is replaced by the quadratic form
domain Q(A) in the deﬁnition of U. In addition, as long as E
n
is an eigen
value, the sup and inf are in fact max and min, explaining the name.
Corollary 4.10. Suppose A and B are selfadjoint operators with A ≥ B
(i.e. A−B ≥ 0), then E
n
(A) ≥ E
n
(B).
Problem 4.2. Suppose A, A
n
are bounded and A
n
→ A. Then E
k
(A
n
) →
E
k
(A). (Hint A−A
n
 ≤ ε is equivalent to A−ε ≤ A ≤ A+ε.)
4.4. Estimating eigenspaces
Next, we show that the dimension of the range of P
A
(Ω) can be estimated
if we have some functions which lie approximately in this space.
Theorem 4.11. Suppose A is a selfadjoint operator and ψ
j
, 1 ≤ j ≤ k,
are linearly independent elements of a H.
(i). Let λ ∈ R, ψ
j
∈ Q(A). If
'ψ, Aψ` < λψ
2
(4.38)
for any nonzero linear combination ψ =
¸
k
j=1
c
j
ψ
j
, then
dimRan P
A
((−∞, λ)) ≥ k. (4.39)
Similarly, 'ψ, Aψ` > λψ
2
implies dimRan P
A
((λ, ∞)) ≥ k.
(ii). Let λ
1
< λ
2
, ψ
j
∈ D(A). If
(A−
λ
2
+λ
1
2
)ψ <
λ
2
−λ
1
2
ψ (4.40)
for any nonzero linear combination ψ =
¸
k
j=1
c
j
ψ
j
, then
dimRan P
A
((λ
1
, λ
2
)) ≥ k. (4.41)
4.5. Tensor products of operators 105
Proof. (i). Let M = span¦ψ
j
¦ ⊆ H. We claim dimP
A
((−∞, λ))M =
dimM = k. For this it suﬃces to show KerP
A
((−∞, λ))[
M
= ¦0¦. Sup
pose P
A
((−∞, λ))ψ = 0, ψ = 0. Then we see that for any nonzero linear
combination ψ
'ψ, Aψ` =
R
η dµ
ψ
(η) =
[λ,∞)
η dµ
ψ
(η)
≥ λ
[λ,∞)
dµ
ψ
(η) = λψ
2
. (4.42)
This contradicts our assumption (4.38).
(ii). Using the same notation as before we need to show KerP
A
((λ
1
, λ
2
))[
M
=
¦0¦. If P
A
((λ
1
, λ
2
))ψ = 0, ψ = 0, then,
(A−
λ
2
+λ
1
2
)ψ
2
=
R
(x −
λ
2
+λ
1
2
)
2
dµ
ψ
(x) =
Ω
x
2
dµ
ψ
(x +
λ
2
+λ
1
2
)
≥
(λ
2
−λ
1
)
2
4
Ω
dµ
ψ
(x +
λ
2
+λ
1
2
) =
(λ
2
−λ
1
)
2
4
ψ
2
, (4.43)
where Ω = (−∞, −(λ
2
−λ
1
)/2]∪[(λ
2
−λ
1
)/2, ∞). But this is a contradiction
as before.
4.5. Tensor products of operators
Suppose A
j
, 1 ≤ j ≤ n, are selfadjoint operators on H
j
. For every monomial
λ
n
1
1
λ
nn
n
we can deﬁne
(A
n
1
1
⊗ ⊗A
nn
n
)ψ
1
⊗ ⊗ψ
n
= (A
n
1
1
ψ
1
) ⊗ ⊗(A
nn
n
ψ
n
), ψ
j
∈ D(A
n
j
j
).
(4.44)
Hence for every polynomial P(λ
1
, . . . , λ
n
) of degree N we can deﬁne
P(A
1
, . . . , A
n
)ψ
1
⊗ ⊗ψ
n
, ψ
j
∈ D(A
N
j
), (4.45)
and extend this deﬁnition to obtain a linear operator on the set
D = span¦ψ
1
⊗ ⊗ψ
n
[ ψ
j
∈ D(A
N
j
)¦. (4.46)
Moreover, if P is realvalued, then the operator P(A
1
, . . . , A
n
) on D is sym
metric and we can consider its closure, which will again be denoted by
P(A
1
, . . . , A
n
).
Theorem 4.12. Suppose A
j
, 1 ≤ j ≤ n, are selfadjoint operators on H
j
and let P(λ
1
, . . . , λ
n
) be a realvalued polynomial and deﬁne P(A
1
, . . . , A
n
)
as above.
Then P(A
1
, . . . , A
n
) is selfadjoint and its spectrum is the closure of the
range of P on the product of the spectra of the A
j
, that is,
σ(P(A
1
, . . . , A
n
)) = P(σ(A
1
), . . . , σ(A
n
)). (4.47)
106 4. Applications of the spectral theorem
Proof. By the spectral theorem it is no restriction to assume that A
j
is
multiplication by λ
j
on L
2
(R, dµ
j
) and P(A
1
, . . . , A
n
) is hence multiplication
by P(λ
1
, . . . , λ
n
) on L
2
(R
n
, dµ
1
dµ
n
). Since D contains the set of
all functions ψ
1
(λ
1
) ψ
n
(λ
n
) for which ψ
j
∈ L
2
c
(R, dµ
j
) it follows that the
domain of the closure of P contains L
2
c
(R
n
, dµ
1
dµ
n
). Hence P is
the maximally deﬁned multiplication operator by P(λ
1
, . . . , λ
n
), which is
selfadjoint.
Now let λ = P(λ
1
, . . . , λ
n
) with λ
j
∈ σ(A
j
). Then there exists Weyl
sequences ψ
j,k
∈ D(A
N
j
) with (A
j
− λ
j
)ψ
j,k
→ 0 as k → ∞. Then, (P −
λ)ψ
k
→ 0, where ψ
k
= ψ
1,k
⊗ ⊗ ψ
1,k
and hence λ ∈ σ(P). Conversely,
if λ ∈ P(σ(A
1
), . . . , σ(A
n
)), then [P(λ
1
, . . . , λ
n
) − λ[ ≥ ε for a.e. λ
j
with
respect to µ
j
and hence (P−λ)
−1
exists and is bounded, that is λ ∈ ρ(P).
The two main cases of interest are A
1
⊗A
2
, in which case
σ(A
1
⊗A
2
) = σ(A
1
)σ(A
2
) = ¦λ
1
λ
2
[λ
j
∈ σ(A
j
)¦, (4.48)
and A
1
⊗I +I ⊗A
2
, in which case
σ(A
1
⊗I +I ⊗A
2
) = σ(A
1
) +σ(A
2
) = ¦λ
1
+λ
2
[λ
j
∈ σ(A
j
)¦. (4.49)
Chapter 5
Quantum dynamics
As in the ﬁnite dimensional case, the solution of the Schr¨odinger equation
i
d
dt
ψ(t) = Hψ(t) (5.1)
is given by
ψ(t) = exp(−itH)ψ(0). (5.2)
A detailed investigation of this formula will be our ﬁrst task. Moreover, in
the ﬁnite dimensional case the dynamics is understood once the eigenvalues
are known and the same is true in our case once we know the spectrum. Note
that, like any Hamiltonian system from classical mechanics, our system is
not hyperbolic (i.e., the spectrum is not away from the real axis) and hence
simple results like, all solutions tend to the equilibrium position cannot be
expected.
5.1. The time evolution and Stone’s theorem
In this section we want to have a look at the initial value problem associated
with the Schr¨odinger equation (2.12) in the Hilbert space H. If H is one
dimensional (and hence A is a real number), the solution is given by
ψ(t) = e
−itA
ψ(0). (5.3)
Our hope is that this formula also applies in the general case and that we
can reconstruct a oneparameter unitary group U(t) from its generator A
(compare (2.11)) via U(t) = exp(−itA). We ﬁrst investigate the family of
operators exp(−itA).
Theorem 5.1. Let A be selfadjoint and let U(t) = exp(−itA).
(i). U(t) is a strongly continuous oneparameter unitary group.
107
108 5. Quantum dynamics
(ii). The limit lim
t→0
1
t
(U(t)ψ − ψ) exists if and only if ψ ∈ D(A) in
which case lim
t→0
1
t
(U(t)ψ −ψ) = −iAψ.
(iii). U(t)D(A) = D(A) and AU(t) = U(t)A.
Proof. The group property (i) follows directly from Theorem 3.1 and the
corresponding statements for the function exp(−itλ). To prove strong con
tinuity observe that
lim
t→t
0
e
−itA
ψ −e
−it
0
A
ψ
2
= lim
t→t
0
R
[e
−itλ
−e
−it
0
λ
[
2
dµ
ψ
(λ)
=
R
lim
t→t
0
[e
−itλ
−e
−it
0
λ
[
2
dµ
ψ
(λ) = 0 (5.4)
by the dominated convergence theorem.
Similarly, if ψ ∈ D(A) we obtain
lim
t→0

1
t
(e
−itA
ψ −ψ) + iAψ
2
= lim
t→0
R
[
1
t
(e
−itλ
−1) + iλ[
2
dµ
ψ
(λ) = 0 (5.5)
since [e
itλ
−1[ ≤ [tλ[. Now let
˜
A be the generator deﬁned as in (2.11). Then
˜
A is a symmetric extension of A since we have
'ϕ,
˜
Aψ` = lim
t→0
'ϕ,
i
t
(U(t) −1)ψ` = lim
t→0
'
i
−t
(U(−t) −1)ϕ, ψ` = '
˜
Aϕ, ψ` (5.6)
and hence
˜
A = A by Corollary 2.2. This settles (ii).
To see (iii) replace ψ → U(s)ψ in (ii).
For our original problem this implies that formula (5.3) is indeed the
solution to the initial value problem of the Schr¨odinger equation. Moreover,
'U(t)ψ, AU(t)ψ` = 'U(t)ψ, U(t)Aψ` = 'ψ, Aψ` (5.7)
shows that the expectations of A are time independent. This corresponds
to conservation of energy.
On the other hand, the generator of the time evolution of a quantum
mechanical system should always be a selfadjoint operator since it corre
sponds to an observable (energy). Moreover, there should be a one to one
correspondence between the unitary group and its generator. This is ensured
by Stone’s theorem.
Theorem 5.2 (Stone). Let U(t) be a weakly continuous oneparameter uni
tary group. Then its generator A is selfadjoint and U(t) = exp(−itA).
Proof. First of all observe that weak continuity together with Lemma 1.11 (iv)
shows that U(t) is in fact strongly continuous.
5.1. The time evolution and Stone’s theorem 109
Next we show that A is densely deﬁned. Pick ψ ∈ H and set
ψ
τ
=
τ
0
U(t)ψdt (5.8)
(the integral is deﬁned as in Section 4.1) implying lim
τ→0
τ
−1
ψ
τ
= ψ. More
over,
1
t
(U(t)ψ
τ
−ψ
τ
) =
1
t
t+τ
t
U(s)ψds −
1
t
τ
0
U(s)ψds
=
1
t
τ+t
τ
U(s)ψds −
1
t
t
0
U(s)ψds
=
1
t
U(τ)
t
0
U(s)ψds −
1
t
t
0
U(s)ψds → U(τ)ψ −ψ (5.9)
as t → 0 shows ψ
τ
∈ D(A). As in the proof of the previous theorem, we can
show that A is symmetric and that U(t)D(A) = D(A).
Next, let us prove that A is essentially selfadjoint. By Lemma 2.6 it
suﬃces to prove Ker(A
∗
−z
∗
) = ¦0¦ for z ∈ C`R. Suppose A
∗
ϕ = z
∗
ϕ, then
for each ψ ∈ D(A) we have
d
dt
'ϕ, U(t)ψ` = 'ϕ, −iAU(t)ψ` = −i'A
∗
ϕ, U(t)ψ`
= −iz'ϕ, U(t)ψ` (5.10)
and hence 'ϕ, U(t)ψ` = exp(−izt)'ϕ, ψ`. Since the left hand side is bounded
for all t ∈ R and the exponential on the right hand side is not, we must have
'ϕ, ψ` = 0 implying ϕ = 0 since D(A) is dense.
So A is essentially selfadjoint and we can introduce V (t) = exp(−itA).
We are done if we can show U(t) = V (t).
Let ψ ∈ D(A) and abbreviate ψ(t) = (U(t) −V (t))ψ. Then
lim
s→0
ψ(t +s) −ψ(t)
s
= iAψ(t) (5.11)
and hence
d
dt
ψ(t)
2
= 2Re'ψ(t), iAψ(t)` = 0. Since ψ(0) = 0 we have
ψ(t) = 0 and hence U(t) and V (t) coincide on D(A). Furthermore, since
D(A) is dense we have U(t) = V (t) by continuity.
As an immediate consequence of the proof we also note the following
useful criterion.
Corollary 5.3. Suppose D ⊆ D(A) is dense and invariant under U(t).
Then A is essentially selfadjoint on D.
Proof. As in the above proof it follows 'ϕ, ψ` = 0 for any ϕ ∈ Ker(A
∗
−z
∗
)
and ψ ∈ D.
110 5. Quantum dynamics
Note that by Lemma 4.8 two strongly continuous oneparameter groups
commute
[e
−itA
, e
−isB
] = 0 (5.12)
if and only if the generators commute.
Clearly, for a physicist, one of the goals must be to understand the time
evolution of a quantum mechanical system. We have seen that the time
evolution is generated by a selfadjoint operator, the Hamiltonian, and is
given by a linear ﬁrst order diﬀerential equation, the Schr¨odinger equation.
To understand the dynamics of such a ﬁrst order diﬀerential equation, one
must understand the spectrum of the generator. Some general tools for this
endeavor will be provided in the following sections.
Problem 5.1. Let H = L
2
(0, 2π) and consider the one parameter unitary
group given by U(t)f(x) = f(x −t mod 2π). What is the generator of U?
5.2. The RAGE theorem
Now, let us discuss why the decomposition of the spectrum introduced in
Section 3.3 is of physical relevance. Let ϕ = ψ = 1. The vector 'ϕ, ψ`ϕ
is the projection of ψ onto the (onedimensional) subspace spanned by ϕ.
Hence ['ϕ, ψ`[
2
can be viewed as the part of ψ which is in the state ϕ. A
ﬁrst question one might rise is, how does
['ϕ, U(t)ψ`[
2
, U(t) = e
−itA
, (5.13)
behave as t → ∞? By the spectral theorem,
ˆ µ
ϕ,ψ
(t) = 'ϕ, U(t)ψ` =
R
e
−itλ
dµ
ϕ,ψ
(λ) (5.14)
is the Fourier transform of the measure µ
ϕ,ψ
. Thus our question is an
swered by Wiener’s theorem.
Theorem 5.4 (Wiener). Let µ be a ﬁnite complex Borel measure on R and
let
ˆ µ(t) =
R
e
−itλ
dµ(λ) (5.15)
be its Fourier transform. Then the Ces`aro time average of ˆ µ(t) has the
following limit
lim
T→∞
1
T
T
0
[ˆ µ(t)[
2
dt =
¸
λ∈R
[µ(¦λ¦)[
2
, (5.16)
where the sum on the right hand side is ﬁnite.
5.2. The RAGE theorem 111
Proof. By Fubini we have
1
T
T
0
[ˆ µ(t)[
2
dt =
1
T
T
0
R
R
e
−i(x−y)t
dµ(x)dµ
∗
(y)dt
=
R
R
1
T
T
0
e
−i(x−y)t
dt
dµ(x)dµ
∗
(y). (5.17)
The function in parentheses is bounded by one and converges pointwise to
χ
{0}
(x − y) as T → ∞. Thus, by the dominated convergence theorem, the
limit of the above expression is given by
R
R
χ
{0}
(x −y)dµ(x)dµ
∗
(y) =
R
µ(¦y¦)dµ
∗
(y) =
¸
y∈R
[µ(¦y¦)[
2
. (5.18)
To apply this result to our situation, observe that the subspaces H
ac
,
H
sc
, and H
pp
are invariant with respect to time evolution since P
xx
U(t) =
χ
Mxx
(A) exp(−itA) = exp(−itA)χ
Mxx
(A) = U(t)P
xx
, xx ∈ ¦ac, sc, pp¦.
Moreover, if ψ ∈ H
xx
we have P
xx
ψ = ψ which shows 'ϕ, f(A)ψ` =
'ϕ, P
xx
f(A)ψ` = 'P
xx
ϕ, f(A)ψ` implying dµ
ϕ,ψ
= dµ
P
xx
ϕ,ψ
. Thus if µ
ψ
is ac, sc, or pp, so is µ
ϕ,ψ
for every ϕ ∈ H.
That is, if ψ ∈ H
c
= H
ac
⊕H
sc
, then the Ces`aro mean of 'ϕ, U(t)ψ` tends
to zero. In other words, the average of the probability of ﬁnding the system
in any prescribed state tends to zero if we start in the continuous subspace
H
c
of A.
If ψ ∈ H
ac
, then dµ
ϕ,ψ
is absolutely continuous with respect to Lebesgue
measure and thus ˆ µ
ϕ,ψ
(t) is continuous and tends to zero as [t[ → ∞. In
fact, this follows from the RiemannLebesgue lemma (see Lemma 7.6 below).
Now we want to draw some additional consequences from Wiener’s the
orem. This will eventually yield a dynamical characterization of the contin
uous and pure point spectrum due to Ruelle, Amrein, Gorgescu, and Enß.
But ﬁrst we need a few deﬁnitions.
An operator K ∈ L(H) is called ﬁnite rank if its range is ﬁnite dimen
sional. The dimension n = dimRan(K) is called the rank of K. If ¦ψ
j
¦
n
j=1
is an orthonormal basis for Ran(K) we have
Kψ =
n
¸
j=1
'ψ
j
, Kψ`ψ
j
=
n
¸
j=1
'ϕ
j
, ψ`ψ
j
, (5.19)
112 5. Quantum dynamics
where ϕ
j
= K
∗
ψ
j
. The elements ϕ
j
are linearly independent since Ran(K) =
Ker(K
∗
)
⊥
. Hence every ﬁnite rank operator is of the form (5.19). In addi
tion, the adjoint of K is also ﬁnite rank and given by
K
∗
ψ =
n
¸
j=1
'ψ
j
, ψ`ϕ
j
. (5.20)
The closure of the set of all ﬁnite rank operators in L(H) is called the set
of compact operators C(H). It is straightforward to verify (Problem 5.2)
Lemma 5.5. The set of all compact operators C(H) is a closed ∗ideal in
L(H).
There is also a weaker version of compactness which is useful for us. The
operator K is called relatively compact with respect to A if
KR
A
(z) ∈ C(H) (5.21)
for one z ∈ ρ(A). By the ﬁrst resolvent identity this then follows for all
z ∈ ρ(A). In particular we have D(A) ⊆ D(K).
Now let us return to our original problem.
Theorem 5.6. Let A be selfadjoint and suppose K is relatively compact.
Then
lim
T→∞
1
T
T
0
Ke
−itA
P
c
ψ
2
dt = 0 and lim
t→∞
Ke
−itA
P
ac
ψ = 0
(5.22)
for every ψ ∈ D(A). In particular, if K is also bounded, then the result
holds for any ψ ∈ H.
Proof. Let ψ ∈ H
c
respectively ψ ∈ H
ac
and drop the projectors. Then, if K
is a rank one operator (i.e., K = 'ϕ
1
, .`ϕ
2
), the claim follows from Wiener’s
theorem respectively the RiemannLebesgue lemma. Hence it holds for any
ﬁnite rank operator K.
If K is compact, there is a sequence K
n
of ﬁnite rank operators such
that K −K
n
 ≤ 1/n and hence
Ke
−itA
ψ ≤ K
n
e
−itA
ψ +
1
n
ψ. (5.23)
Thus the claim holds for any compact operator K.
If ψ ∈ D(A) we can set ψ = (A − i)
−1
ϕ, where ϕ ∈ H
c
if and only if
ψ ∈ H
c
(since H
c
reduces A). Since K(A + i)
−1
is compact by assumption,
the claim can be reduced to the previous situation. If, in addition, K is
bounded, we can ﬁnd a sequence ψ
n
∈ D(A) such that ψ −ψ
n
 ≤ 1/n and
hence
Ke
−itA
ψ ≤ Ke
−itA
ψ
n
 +
1
n
K, (5.24)
5.2. The RAGE theorem 113
concluding the proof.
With the help of this result we can now prove an abstract version of the
RAGE theorem.
Theorem 5.7 (RAGE). Let A be selfadjoint. Suppose K
n
∈ L(H) is a se
quence of relatively compact operators which converges strongly to the iden
tity. Then
H
c
= ¦ψ ∈ H[ lim
n→∞
lim
T→∞
1
T
T
0
K
n
e
−itA
ψdt = 0¦,
H
pp
= ¦ψ ∈ H[ lim
n→∞
sup
t≥0
(I −K
n
)e
−itA
ψ = 0¦. (5.25)
Proof. Abbreviate ψ(t) = exp(−itA)ψ. We begin with the ﬁrst equation.
Let ψ ∈ H
c
, then
1
T
T
0
K
n
ψ(t)dt ≤
1
T
T
0
K
n
ψ(t)
2
dt
1/2
→ 0 (5.26)
by CauchySchwarz and the previous theorem. Conversely, if ψ ∈ H
c
we
can write ψ = ψ
c
+ ψ
pp
. By our previous estimate it suﬃces to show
K
n
ψ
pp
(t) ≥ ε > 0 for n large. In fact, we even claim
lim
n→∞
sup
t≥0
K
n
ψ
pp
(t) −ψ
pp
(t) = 0. (5.27)
By the spectral theorem, we can write ψ
pp
(t) =
¸
j
α
j
(t)ψ
j
, where the ψ
j
are orthonormal eigenfunctions and α
j
(t) = exp(−itλ
j
)α
j
. Truncate this
expansion after N terms, then this part converges uniformly to the desired
limit by strong convergence of K
n
. Moreover, by Lemma 1.13 we have
K
n
 ≤ M, and hence the error can be made arbitrarily small by choosing
N large.
Now let us turn to the second equation. If ψ ∈ H
pp
the claim follows
by (5.27). Conversely, if ψ ∈ H
pp
we can write ψ = ψ
c
+ ψ
pp
and by our
previous estimate it suﬃces to show that (I −K
n
)ψ
c
(t) does not tend to
0 as n → ∞. If it would, we would have
0 = lim
T→∞
1
T
T
0
(I −K
n
)ψ
c
(t)
2
dt
≥ ψ
c
(t)
2
− lim
T→∞
1
T
T
0
K
n
ψ
c
(t)
2
dt = ψ
c
(t)
2
, (5.28)
a contradiction.
In summary, regularity properties of spectral measures are related to
the long time behavior of the corresponding quantum mechanical system.
114 5. Quantum dynamics
However, a more detailed investigation of this topic is beyond the scope of
this manuscript. For a survey containing several recent results see [9].
It is often convenient to treat the observables as time dependent rather
than the states. We set
K(t) = e
itA
Ke
−itA
(5.29)
and note
'ψ(t), Kψ(t)` = 'ψ, K(t)ψ`, ψ(t) = e
−itA
ψ. (5.30)
This point of view is often referred to as Heisenberg picture in the physics
literature. If K is unbounded we will assume D(A) ⊆ D(K) such that the
above equations make sense at least for ψ ∈ D(A). The main interest is
the behavior of K(t) for large t. The strong limits are called asymptotic
observables if they exist.
Theorem 5.8. Suppose A is selfadjoint and K is relatively compact. Then
lim
T→∞
1
T
T
0
e
itA
Ke
−itA
ψdt =
¸
λ∈σp(A)
P
A
(¦λ¦)KP
A
(¦λ¦)ψ, ψ ∈ D(A).
(5.31)
If K is in addition bounded, the result holds for any ψ ∈ H.
Proof. We will assume that K is bounded. To obtain the general result,
use the same trick as before and replace K by KR
A
(z). Write ψ = ψ
c
+ψ
pp
.
Then
lim
T→∞
1
T

T
0
K(t)ψ
c
dt ≤ lim
T→∞
1
T
T
0
K(t)ψ
c
dt = 0 (5.32)
by Theorem 5.6. As in the proof of the previous theorem we can write
ψ
pp
=
¸
j
α
j
ψ
j
and hence
¸
j
α
j
1
T
T
0
K(t)ψ
j
dt =
¸
j
α
j
1
T
T
0
e
it(A−λ
j
)
dt
Kψ
j
. (5.33)
As in the proof of Wiener’s theorem, we see that the operator in parenthesis
tends to P
A
(¦λ
j
¦) strongly as T → ∞. Since this operator is also bounded
by 1 for all T, we can interchange the limit with the summation and the
claim follows.
We also note the following corollary.
Corollary 5.9. Under the same assumptions as in the RAGE theorem we
have
lim
n→∞
lim
T→∞
1
T
T
0
e
itA
K
n
e
−itA
ψdt = P
pp
ψ (5.34)
5.3. The Trotter product formula 115
respectively
lim
n→∞
lim
T→∞
1
T
T
0
e
−itA
(I −K
n
)e
−itA
ψdt = P
c
ψ. (5.35)
Problem 5.2. Prove Lemma 5.5.
Problem 5.3. Prove Corollary 5.9.
5.3. The Trotter product formula
In many situations the operator is of the form A + B, where e
itA
and e
itB
can be computed explicitly. Since A and B will not commute in general, we
cannot obtain e
it(A+B)
from e
itA
e
itB
. However, we at least have:
Theorem 5.10 (Trotter product formula). Suppose, A, B, and A+B are
selfadjoint. Then
e
it(A+B)
= slim
n→∞
e
i
t
n
A
e
i
t
n
B
n
. (5.36)
Proof. First of all note that we have
e
iτA
e
iτB
n
−e
it(A+B)
=
n−1
¸
j=0
e
iτA
e
iτB
n−1−j
e
iτA
e
iτB
−e
iτ(A+B)
e
iτ(A+B)
j
,(5.37)
where τ =
t
n
, and hence
(e
iτA
e
iτB
)
n
−e
it(A+B)
ψ ≤ [t[ max
s≤t
F
τ
(s), (5.38)
where
F
τ
(s) = 
1
τ
(e
iτA
e
iτB
−e
iτ(A+B)
)e
is(A+B)
ψ. (5.39)
Now for ψ ∈ D(A+B) = D(A) ∩ D(B) we have
1
τ
(e
iτA
e
iτB
−e
iτ(A+B)
)ψ → iAψ + iBψ −i(A+B)ψ = 0 (5.40)
as τ → 0. So lim
τ→0
F
τ
(s) = 0 at least pointwise, but we need this uniformly
with respect to s ∈ [−[t[, [t[].
Pointwise convergence implies

1
τ
(e
iτA
e
iτB
−e
iτ(A+B)
)ψ ≤ C(ψ) (5.41)
and, since D(A+B) is a Hilbert space when equipped with the graph norm
ψ
2
Γ(A+B)
= ψ
2
+(A+B)ψ
2
, we can invoke the uniform boundedness
principle to obtain

1
τ
(e
iτA
e
iτB
−e
iτ(A+B)
)ψ ≤ Cψ
Γ(A+B)
. (5.42)
116 5. Quantum dynamics
Now
[F
τ
(s) −F
τ
(r)[ ≤ 
1
τ
(e
iτA
e
iτB
−e
iτ(A+B)
)(e
is(A+B)
−e
ir(A+B)
)ψ
≤ C(e
is(A+B)
−e
ir(A+B)
)ψ
Γ(A+B)
. (5.43)
shows that F
τ
(.) is uniformly continuous and the claim follows by a standard
ε
2
argument.
If the operators are semibounded from below the same proof shows
Theorem 5.11 (Trotter product formula). Suppose, A, B, and A+B are
selfadjoint and semibounded from below. Then
e
−t(A+B)
= slim
n→∞
e
−
t
n
A
e
−
t
n
B
n
, t ≥ 0. (5.44)
Problem 5.4. Proof Theorem 5.11.
Chapter 6
Perturbation theory for
selfadjoint operators
The Hamiltonian of a quantum mechanical system is usually the sum of
the kinetic energy H
0
(free Schr¨odinger operator) plus an operator V cor
responding to the potential energy. Since H
0
is easy to investigate, one
usually tries to consider V as a perturbation of H
0
. This will only work
if V is small with respect to H
0
. Hence we study such perturbations of
selfadjoint operators next.
6.1. Relatively bounded operators and the
Kato–Rellich theorem
An operator B is called A bounded or relatively bounded with respect
to A if D(A) ⊆ D(B) and if there are constants a, b ≥ 0 such that
Bψ ≤ aAψ +bψ, ψ ∈ D(A). (6.1)
The inﬁmum of all constants a for which a corresponding b exists such that
(6.1) holds is called the Abound of B.
The triangle inequality implies
Lemma 6.1. Suppose B
j
, j = 1, 2, are A bounded with respective Abounds
a
i
, i = 1, 2. Then α
1
B
1
+ α
2
B
2
is also A bounded with Abound less than
[α
1
[a
1
+ [α
2
[a
2
. In particular, the set of all A bounded operators forms a
linear space.
There are also the following equivalent characterizations:
117
118 6. Perturbation theory for selfadjoint operators
Lemma 6.2. Suppose A is closed and B is closable. Then the following are
equivalent:
(i) B is A bounded.
(ii) D(A) ⊆ D(B).
(iii) BR
A
(z) is bounded for one (and hence for all) z ∈ ρ(A).
Moreover, the Abound of B is no larger then inf
z∈ρ(A)
BR
A
(z).
Proof. (i) ⇒ (ii) is true by deﬁnition. (ii) ⇒ (iii) since BR
A
(z) is a closed
(Problem 2.6) operator deﬁned on all of H and hence bounded by the closed
graph theorem (Theorem 2.7). To see (iii) ⇒ (i) let ψ ∈ D(A), then
Bψ = BR
A
(z)(A−z)ψ ≤ a(A−z)ψ ≤ aAψ + (a[z[)ψ, (6.2)
where a = BR
A
(z). Finally, note that if BR
A
(z) is bounded for one
z ∈ ρ(A), it is bounded for all z ∈ ρ(A) by the ﬁrst resolvent formula.
We are mainly interested in the situation where A is selfadjoint and B
is symmetric. Hence we will restrict our attention to this case.
Lemma 6.3. Suppose A is selfadjoint and B relatively bounded. The A
bound of B is given by
lim
λ→∞
BR
A
(±iλ). (6.3)
If A is bounded from below, we can also replace ±iλ by −λ.
Proof. Let ϕ = R
A
(±iλ)ψ, λ > 0, and let a
∞
be the Abound of B. If B
is A bounded, then (use the spectral theorem to estimate the norms)
BR
A
(±iλ)ψ ≤ aAR
A
(±iλ)ψ +bR
A
(±iλ)ψ ≤ (a +
b
λ
)ψ. (6.4)
Hence limsup
λ
BR
A
(±iλ) ≤ a
∞
which, together with a
∞
≤ inf
λ
BR
A
(±iλ)
from the previous lemma, proves the claim.
The case where A is bounded from below is similar.
Now we will show the basic perturbation result due to Kato and Rellich.
Theorem 6.4 (Kato–Rellich). Suppose A is (essentially) selfadjoint and
B is symmetric with Abound less then one. Then A + B, D(A + B) =
D(A), is (essentially) selfadjoint. If A is essentially selfadjoint we have
D(A) ⊆ D(B) and A+B = A+B.
If A is bounded from below by γ, then A + B is bounded from below by
min(γ, b/(a −1)).
6.2. More on compact operators 119
Proof. Since D(A) ⊆ D(B) and D(A) ⊆ D(A+B) by (6.1) we can assume
that A is closed (i.e., selfadjoint). It suﬃces to show that Ran(A+B±iλ) =
H. By the above lemma we can ﬁnd a λ > 0 such that BR
A
(±iλ) < 1.
Hence −1 ∈ ρ(BR
A
(±iλ)) and thus I +BR
A
(±iλ) is onto. Thus
(A+B ±iλ) = (I +BR
A
(±iλ))(A±iλ) (6.5)
is onto and the proof of the ﬁrst part is complete.
If A is bounded from below we can replace ±iλ by −λ and the above
equation shows that R
A+B
exists for λ suﬃciently large. By the proof of
the previous lemma we can choose −λ < min(γ, b/(a −1)).
Finally, let us show that there is also a connection between the resolvents.
Lemma 6.5. Suppose A and B are closed and D(A) ⊆ D(B). Then we
have the second resolvent formula
R
A+B
(z) −R
A
(z) = −R
A
(z)BR
A+B
(z) = −R
A+B
(z)BR
A
(z) (6.6)
for z ∈ ρ(A) ∩ ρ(A+B).
Proof. We compute
R
A+B
(z) +R
A
(z)BR
A+B
(z) = R
A
(z)(A+B −z)R
A+B
(z) = R
A
(z). (6.7)
The second identity is similar.
Problem 6.1. Compute the resolvent of A+α'ψ, .`ψ. (Hint: Show
(I +α'ϕ, .`ψ)
−1
= I −
α
1 +α'ϕ, ψ`
'ϕ, .`ψ (6.8)
and use the second resolvent formula.)
6.2. More on compact operators
Recall from Section 5.2 that we have introduced the set of compact operators
C(H) as the closure of the set of all ﬁnite rank operators in L(H). Before we
can proceed, we need to establish some further results for such operators.
We begin by investigating the spectrum of selfadjoint compact operators
and show that the spectral theorem takes a particular simple form in this
case.
We introduce some notation ﬁrst. The discrete spectrum σ
d
(A) is the
set of all eigenvalues which are discrete points of the spectrum and whose
corresponding eigenspace is ﬁnite dimensional. The complement of the dis
crete spectrum is called the essential spectrum σ
ess
(A) = σ(A)`σ
d
(A).
If A is selfadjoint we might equivalently set
σ
d
(A) = ¦λ ∈ σ
p
(A)[rank(P
A
((λ −ε, λ +ε))) < ∞ for some ε > 0¦. (6.9)
120 6. Perturbation theory for selfadjoint operators
respectively
σ
ess
(A) = ¦λ ∈ R[rank(P
A
((λ −ε, λ +ε))) = ∞ for all ε > 0¦. (6.10)
Theorem 6.6 (Spectral theorem for compact operators). Suppose the op
erator K is selfadjoint and compact. Then the spectrum of K consists of
an at most countable number of eigenvalues which can only cluster at 0.
Moreover, the eigenspace to each nonzero eigenvalue is ﬁnite dimensional.
In other words,
σ
ess
(K) ⊆ ¦0¦, (6.11)
where equality holds if and only if H is inﬁnite dimensional.
In addition, we have
K =
¸
λ∈σ(K)
λP
K
(¦λ¦). (6.12)
Proof. It suﬃces to show rank(P
K
((λ −ε, λ +ε))) < ∞ for 0 < ε < [λ[.
Let K
n
be a sequence of ﬁnite rank operators such that K−K
n
 ≤ 1/n.
If RanP
K
((λ−ε, λ+ε)) is inﬁnite dimensional we can ﬁnd a vector ψ
n
in this
range such that ψ
n
 = 1 and K
n
ψ
n
= 0. But this yields a contradiction
1
n
≥ ['ψ
n
, (K −K
n
)ψ
n
`[ = ['ψ
n
, Kψ
n
`[ ≥ [λ[ −ε > 0 (6.13)
by (4.2).
As a consequence we obtain the canonical form of a general compact
operator.
Theorem 6.7 (Canonical form of compact operators). Let K be compact.
There exists orthonormal sets ¦
ˆ
φ
j
¦, ¦φ
j
¦ and positive numbers s
j
= s
j
(K)
such that
K =
¸
j
s
j
'φ
j
, .`
ˆ
φ
j
, K
∗
=
¸
j
s
j
'
ˆ
φ
j
, .`φ
j
. (6.14)
Note Kφ
j
= s
j
ˆ
φ
j
and K
∗
ˆ
φ
j
= s
j
φ
j
, and hence K
∗
Kφ
j
= s
2
j
φ
j
and KK
∗
ˆ
φ
j
=
s
2
j
ˆ
φ
j
.
The numbers s
j
(K)
2
> 0 are the nonzero eigenvalues of KK
∗
respec
tively K
∗
K (counted with multiplicity) and s
j
(K) = s
j
(K
∗
) = s
j
are called
singular values of K. There are either ﬁnitely many singular values (if K
is ﬁnite rank) or they converge to zero.
Proof. By Lemma 5.5 K
∗
K is compact and hence Theorem 6.6 applies.
Let ¦φ
j
¦ be an orthonormal basis of eigenvectors for P
K
∗
K
((0, ∞))H and let
6.2. More on compact operators 121
s
2
j
be the eigenvalue corresponding to φ
j
. Then, for any ψ ∈ H we can write
ψ =
¸
j
'φ
j
, ψ`φ
j
+
˜
ψ (6.15)
with
˜
ψ ∈ Ker(K
∗
K) = Ker(K). Then
Kψ =
¸
j
s
j
'φ
j
, ψ`
ˆ
φ
j
, (6.16)
where
ˆ
φ
j
= s
−1
j
Kφ
j
, since K
˜
ψ
2
= '
˜
ψ, K
∗
K
˜
ψ` = 0. By '
ˆ
φ
j
,
ˆ
φ
k
` =
(s
j
s
k
)
−1
'Kφ
j
, Kφ
k
` = (s
j
s
k
)
−1
'K
∗
Kφ
j
, φ
k
` = s
j
s
−1
k
'φ
j
, φ
k
` we see that
¦
ˆ
φ
j
¦ are orthonormal and the formula for K
∗
follows by taking the adjoint
of the formula for K (Problem 6.2).
If K is selfadjoint then φ
j
= σ
j
ˆ
φ
j
, σ
2
j
= 1 are the eigenvectors of K and
σ
j
s
j
are the corresponding eigenvalues.
Moreover, note that we have
K = max
j
s
j
(K). (6.17)
Finally, let me remark that there are a number of other equivalent deﬁ
nitions for compact operators.
Lemma 6.8. For K ∈ L(H) the following statements are equivalent:
(i) K is compact.
(i’) K
∗
is compact.
(ii) A
n
∈ L(H) and A
n
s
→ A strongly implies A
n
K → AK.
(iii) ψ
n
ψ weakly implies Kψ
n
→ Kψ in norm.
(iv) ψ
n
bounded implies that Kψ
n
has a (norm) convergent subse
quence.
Proof. (i) ⇔ (i’). This is immediate from Theorem 6.7.
(i) ⇒(ii). Translating A
n
→ A
n
−A it is no restriction to assume A = 0.
Since A
n
 ≤ M it suﬃces to consider the case where K is ﬁnite rank. Then
(by (6.14))
A
n
K
2
= sup
ψ=1
N
¸
j=1
s
j
['φ
j
, ψ`[
2
A
n
ˆ
φ
j

2
≤
N
¸
j=1
s
j
A
n
ˆ
φ
j

2
→ 0. (6.18)
(ii) ⇒ (iii). Again, replace ψ
n
→ ψ
n
− ψ and assume ψ = 0. Choose
A
n
= 'ψ
n
, .`ϕ, ϕ = 1, then Kψ
n

2
= A
n
K
∗
 → 0.
(iii) ⇒ (iv). If ψ
n
is bounded it has a weakly convergent subsequence
by Lemma 1.12. Now apply (iii) to this subsequence.
122 6. Perturbation theory for selfadjoint operators
(iv) ⇒ (i). Let ϕ
j
be an orthonormal basis and set
K
n
=
n
¸
j=1
'ϕ
j
, .`Kϕ
j
. (6.19)
Then
γ
n
= K −K
n
 = sup
ψ∈span{ϕ
j
}
∞
j=n
,ψ=1
Kψ (6.20)
is a decreasing sequence tending to a limit ε ≥ 0. Moreover, we can ﬁnd
a sequence of unit vectors ψ
n
∈ span¦ϕ
j
¦
∞
j=n
for which Kψ
n
 ≥ ε. By
assumption, Kψ
n
has a convergent subsequence which, since ψ
n
converges
weakly to 0, converges to 0. Hence ε must be 0 and we are done.
The last condition explains the name compact. Moreover, note that you
cannot replace A
n
K → AK by KA
n
→ KA unless you additionally require
A
n
to be normal (then this follows by taking adjoints — recall that only
for normal operators taking adjoints is continuous with respect to strong
convergence). Without the requirement that A
n
is normal the claim is wrong
as the following example shows.
Example. Let H =
2
(N), A
n
the operator which shifts each sequence n
places to the left, and K = 'δ
1
, .`δ
1
, where δ
1
= (1, 0, . . . ). Then slimA
n
=
0 but KA
n
 = 1.
Problem 6.2. Deduce the formula for K
∗
from the one for K in (6.14).
6.3. Hilbert–Schmidt and trace class operators
Among the compact operators two special classes or of particular impor
tance. The ﬁrst one are integral operators
Kψ(x) =
M
K(x, y)ψ(y)dµ(y), ψ ∈ L
2
(M, dµ), (6.21)
where K(x, y) ∈ L
2
(M M, dµ ⊕dµ). Such an operator is called Hilbert–
Schmidt operator. Using CauchySchwarz,
M
[Kψ(x)[
2
dµ(x) =
M
M
[K(x, y)ψ(y)[dµ(y)
2
dµ(x)
≤
M
M
[K(x, y)[
2
dµ(y)
M
[ψ(y)[
2
dµ(y)
dµ(x)
=
M
M
[K(x, y)[
2
dµ(y) dµ(x)
M
[ψ(y)[
2
dµ(y)
(6.22)
we see that K is bounded. Next, pick an orthonormal basis ϕ
j
(x) for
L
2
(M, dµ). Then, by Lemma 1.9, ϕ
i
(x)ϕ
j
(y) is an orthonormal basis for
6.3. Hilbert–Schmidt and trace class operators 123
L
2
(M M, dµ ⊕dµ) and
K(x, y) =
¸
i,j
c
i,j
ϕ
i
(x)ϕ
j
(y), c
i,j
= 'ϕ
i
, Kϕ
∗
j
`, (6.23)
where
¸
i,j
[c
i,j
[
2
=
M
M
[K(x, y)[
2
dµ(y) dµ(x) < ∞. (6.24)
In particular,
Kψ(x) =
¸
i,j
c
i,j
'ϕ
∗
j
, ψ`ϕ
i
(x) (6.25)
shows that K can be approximated by ﬁnite rank operators (take ﬁnitely
many terms in the sum) and is hence compact.
Using (6.14) we can also give a diﬀerent characterization of Hilbert–
Schmidt operators.
Lemma 6.9. If H = L
2
(M, dµ), then a compact operator K is Hilbert–
Schmidt if and only if
¸
j
s
j
(K)
2
< ∞ and
¸
j
s
j
(K)
2
=
M
M
[K(x, y)[
2
dµ(x)dµ(y), (6.26)
in this case.
Proof. If K is compact we can deﬁne approximating ﬁnite rank operators
K
n
by considering only ﬁnitely many terms in (6.14):
K
n
=
n
¸
j=1
s
j
'φ
j
, .`
ˆ
φ
j
. (6.27)
Then K
n
has the kernel K
n
(x, y) =
¸
n
j=1
s
j
φ
j
(y)
∗
ˆ
φ
j
(x) and
M
M
[K
n
(x, y)[
2
dµ(x)dµ(y) =
n
¸
j=1
s
j
(K)
2
(6.28)
Now if one side converges, so does the other and in particular, (6.26) holds
in this case.
Hence we will call a compact operator Hilbert–Schmidt if its singular
values satisfy
¸
j
s
j
(K)
2
< ∞. (6.29)
By our lemma this coincides with our previous deﬁnition if H = L
2
(M, dµ).
124 6. Perturbation theory for selfadjoint operators
Since every Hilbert space is isomorphic to some L
2
(M, dµ) we see that
the Hilbert–Schmidt operators together with the norm
K
2
=
¸
j
s
j
(K)
2
1/2
(6.30)
form a Hilbert space (isomorphic to L
2
(MM, dµ⊗dµ)). Note that K
2
=
K
∗

2
(since s
j
(K) = s
j
(K
∗
)). There is another useful characterization for
identifying Hilbert–Schmidt operators:
Lemma 6.10. A compact operator K is Hilbert–Schmidt if and only if
¸
n
Kψ
n

2
< ∞ (6.31)
for some orthonormal set and
¸
n
Kψ
n

2
= K
2
2
(6.32)
in this case.
Proof. This follows from
¸
n
Kψ
n

2
=
¸
n,j
['
ˆ
φ
j
, Kψ
n
`[
2
=
¸
n,j
['K
∗
ˆ
φ
j
, ψ
n
`[
2
=
¸
n
K
∗
ˆ
φ
n

2
=
¸
j
s
j
(K)
2
. (6.33)
Corollary 6.11. The set of Hilbert–Schmidt operators forms a ∗ideal in
L(H) and
KA
2
≤ AK
2
respectively AK
2
≤ AK
2
. (6.34)
Proof. Let K be Hilbert–Schmidt and A bounded. Then AK is compact
and
AK
2
2
=
¸
n
AKψ
n

2
≤ A
2
¸
n
Kψ
n

2
= A
2
K
2
2
. (6.35)
For KA just consider adjoints.
This approach can be generalized by deﬁning
K
p
=
¸
j
s
j
(K)
p
1/p
(6.36)
plus corresponding spaces
.
p
(H) = ¦K ∈ C(H)[K
p
< ∞¦, (6.37)
6.3. Hilbert–Schmidt and trace class operators 125
which are known as Schatten pclass. Note that
K ≤ K
p
. (6.38)
and that by s
j
(K) = s
j
(K
∗
) we have
K
p
= K
∗

p
. (6.39)
Lemma 6.12. The spaces .
p
(H) together with the norm .
p
are Banach
spaces. Moreover,
K
p
= sup
¸
j
['ψ
j
, Kϕ
j
`[
p
1/p
¦ψ
j
¦, ¦ϕ
j
¦ ONS
, (6.40)
where the sup is taken over all orthonormal systems.
Proof. The hard part is to prove (6.40): Choose q such that
1
p
+
1
q
= 1 and
use H¨older’s inequality to obtain (s
j
[...[
2
= (s
p
j
[...[
2
)
1/p
[...[
2/q
)
¸
j
s
j
['ϕ
n
, φ
j
`[
2
≤
¸
j
s
p
j
['ϕ
n
, φ
j
`[
2
1/p
¸
j
['ϕ
n
, φ
j
`[
2
1/q
≤
¸
j
s
p
j
['ϕ
n
, φ
j
`[
2
1/p
. (6.41)
Clearly the analogous equation holds for
ˆ
φ
j
, ψ
n
. Now using CauchySchwarz
we have
['ψ
n
, Kϕ
n
`[
p
=
¸
j
s
1/2
j
'ϕ
n
, φ
j
`s
1/2
j
'
ˆ
φ
j
, ψ
n
`
p
≤
¸
j
s
p
j
['ϕ
n
, φ
j
`[
2
1/2
¸
j
s
p
j
['ψ
n
,
ˆ
φ
j
`[
2
1/2
(6.42)
Summing over n, a second appeal to CauchySchwarz and interchanging the
order of summation ﬁnally gives
¸
n
['ψ
n
, Kϕ
n
`[
p
≤
¸
n,j
s
p
j
['ϕ
n
, φ
j
`[
2
1/2
¸
n,j
s
p
j
['ψ
n
,
ˆ
φ
j
`[
2
1/2
≤
¸
j
s
p
j
1/2
¸
j
s
p
j
1/2
=
¸
j
s
p
j
. (6.43)
Since equality is attained for ϕ
n
= φ
n
and ψ
n
=
ˆ
φ
n
equation (6.40) holds.
126 6. Perturbation theory for selfadjoint operators
Now the rest is straightforward. From
¸
j
['ψ
j
, (K
1
+K
2
)ϕ
j
`[
p
1/p
≤
¸
j
['ψ
j
, K
1
ϕ
j
`[
p
1/p
+
¸
j
['ψ
j
, K
2
ϕ
j
`[
p
1/p
≤ K
1

p
+K
2

p
(6.44)
we infer that .
p
(H) is a vector space and the triangle inequality. The other
requirements for are norm are obvious and it remains to check completeness.
If K
n
is a Cauchy sequence with respect to .
p
it is also a Cauchy sequence
with respect to . (K ≤ K
p
). Since C(H) is closed, there is a compact
K with K −K
n
 → 0 and by K
n

p
≤ C we have
¸
j
['ψ
j
, Kϕ
j
`[
p
1/p
≤ C (6.45)
for any ﬁnite ONS. Since the right hand side is independent of the ONS
(and in particular on the number of vectors), K is in .
p
(H).
The two most important cases are p = 1 and p = 2: .
2
(H) is the space
of Hilbert–Schmidt operators investigated in the previous section and .
1
(H)
is the space of trace class operators. Since Hilbert–Schmidt operators are
easy to identify it is important to relate .
1
(H) with .
2
(H):
Lemma 6.13. An operator is trace class if and only if it can be written as
the product of two Hilbert–Schmidt operators, K = K
1
K
2
, and we have
K
1
≤ K
1

2
K
2

2
(6.46)
in this case.
Proof. By CauchySchwarz we have
¸
n
['ϕ
n
, Kψ
n
`[ =
¸
n
['K
∗
1
ϕ
n
, K
2
ψ
n
`[ ≤
¸
n
K
∗
1
ϕ
n

2
¸
n
K
2
ψ
n

2
1/2
= K
1

2
K
2

2
(6.47)
and hence K = K
1
K
2
is trace calls if both K
1
and K
2
are Hilbert–Schmidt
operators. To see the converse let K be given by (6.14) and choose K
1
=
¸
j
s
j
(K)'φ
j
, .`
ˆ
φ
j
respectively K
2
=
¸
j
s
j
(K)'φ
j
, .`φ
j
.
Corollary 6.14. The set of trace class operators forms a ∗ideal in L(H)
and
KA
1
≤ AK
1
respectively AK
1
≤ AK
1
. (6.48)
6.3. Hilbert–Schmidt and trace class operators 127
Proof. Write K = K
1
K
2
with K
1
, K
2
Hilbert–Schmidt and use Corol
lary 6.11.
Now we can also explain the name trace class:
Lemma 6.15. If K is trace class, then for any ONB ¦ϕ
n
¦ the trace
tr(K) =
¸
n
'ϕ
n
, Kϕ
n
` (6.49)
is ﬁnite and independent of the ONB.
Moreover, the trace is linear and if K
1
≤ K
2
are both trace class we have
tr(K
1
) ≤ tr(K
2
).
Proof. Let ¦ψ
n
¦ be another ONB. If we write K = K
1
K
2
with K
1
, K
2
Hilbert–Schmidt we have
¸
n
'ϕ
n
, K
1
K
2
ϕ
n
` =
¸
n
'K
∗
1
ϕ
n
, K
2
ϕ
n
` =
¸
n,m
'K
∗
1
ϕ
n
, ψ
m
`'ψ
m
, K
2
ϕ
n
`
=
¸
m,n
'K
∗
2
ψ
m
, ϕ
n
`'ϕ
n
, K
1
ψ
m
` =
¸
m
'K
∗
2
ψ
m
, K
1
ψ
m
`
=
¸
m
'ψ
m
, K
2
K
1
ψ
m
`. (6.50)
Hence the trace is independent of the ONB and we even have tr(K
1
K
2
) =
tr(K
2
K
1
).
The rest follows from linearity of the scalar product and K
1
≤ K
2
if and
only if 'ϕ, K
1
ϕ` ≤ 'ϕ, K
2
ϕ` for every ϕ ∈ H.
Clearly for selfadjoint trace class operators, the trace is the sum over
all eigenvalues (counted with their multiplicity). To see this you just have
to choose the ONB to consist of eigenfunctions. This is even true for all
trace class operators and is known as Lidiskij trace theorem (see [17] or [6]
for an easy to read introduction).
Problem 6.3. Let H =
2
(N) and let A be multiplication by a sequence
a(n). Show that A ∈ .
p
(
2
(N)) if and only if a ∈
p
(N). Furthermore, show
that A
p
= a
p
in this case.
Problem 6.4. Show that A ≥ 0 is trace class if (6.49) is ﬁnite for one (and
hence all) ONB. (Hint A is selfadjoint (why?) and A =
√
A
√
A.)
Problem 6.5. Show that for an orthogonal projection P we have
dimRan(P) = tr(P),
where we set tr(P) = ∞ if (6.49) is inﬁnite (for one and hence all ONB by
the previous problem).
128 6. Perturbation theory for selfadjoint operators
Problem 6.6. Show that for K ∈ C we have
[K[ =
¸
j
s
j
'φ
j
, .`φ
j
,
where [K[ =
√
K
∗
K. Conclude that
K
p
= (tr([A[
p
))
1/p
.
Problem 6.7. Show that K :
2
(N) →
2
(N), f(n) →
¸
j∈N
k(n +j)f(j) is
Hilbert–Schmidt if [k(n)[ ≤ C(n), where C(n) is decreasing and summable.
6.4. Relatively compact operators and Weyl’s
theorem
In the previous section we have seen that the sum of a selfadjoint and a sym
metric operator is again selfadjoint if the perturbing operator is small. In
this section we want to study the inﬂuence of perturbations on the spectrum.
Our hope is that at least some parts of the spectrum remain invariant.
Let A be selfadjoint. Note that if we add a multiple of the identity to
A, we shift the entire spectrum. Hence, in general, we cannot expect a (rel
atively) bounded perturbation to leave any part of the spectrum invariant.
Next, if λ
0
is in the discrete spectrum, we can easily remove this eigenvalue
with a ﬁnite rank perturbation of arbitrary small norm. In fact, consider
A+εP
A
(¦λ
0
¦). (6.51)
Hence our only hope is that the remainder, namely the essential spectrum,
is stable under ﬁnite rank perturbations. To show this, we ﬁrst need a good
criterion for a point to be in the essential spectrum of A.
Lemma 6.16 (Weyl criterion). A point λ is in the essential spectrum of
a selfadjoint operator A if and only if there is a sequence ψ
n
such that
ψ
n
 = 1, ψ
n
converges weakly to 0, and (A − λ)ψ
n
 → 0. Moreover, the
sequence can chosen to be orthonormal. Such a sequence is called singular
Weyl sequence.
Proof. Let ψ
n
be a singular Weyl sequence for the point λ
0
. By Lemma 2.12
we have λ
0
∈ σ(A) and hence it suﬃces to show λ
0
∈ σ
d
(A). If λ
0
∈ σ
d
(A) we
can ﬁnd an ε > 0 such that P
ε
= P
A
((λ
0
−ε, λ
0
+ε)) is ﬁnite rank. Consider
˜
ψ
n
= P
ε
ψ
n
. Clearly
˜
ψ
n
converges weakly to zero and (A − λ
0
)
˜
ψ
n
 → 0.
6.4. Relatively compact operators and Weyl’s theorem 129
Moreover,
ψ
n
−
˜
ψ
n

2
=
R\(λ−ε,λ+ε)
dµ
ψn
(λ)
≤
1
ε
2
R\(λ−ε,λ+ε)
(λ −λ
0
)
2
dµ
ψn
(λ)
≤
1
ε
2
(A−λ
0
)ψ
n

2
(6.52)
and hence 
˜
ψ
n
 → 1. Thus ϕ
n
= 
˜
ψ
n

−1
˜
ψ
n
is also a singular Weyl sequence.
But ϕ
n
is a sequence of unit length vectors which lives in a ﬁnite dimensional
space and converges to 0 weakly, a contradiction.
Conversely, if λ
0
∈ σ
ess
(A), consider P
n
= P
A
([λ −
1
n
, λ −
1
n+1
) ∪ (λ +
1
n+1
, λ +
1
n
]). Then rank(P
n
j
) > 0 for an inﬁnite subsequence n
j
. Now pick
ψ
j
∈ RanP
n
j
.
Now let K be a selfadjoint compact operator and ψ
n
a singular Weyl
sequence for A. Then ψ
n
converges weakly to zero and hence
(A+K −λ)ψ
n
 ≤ (A−λ)ψ
n
 +Kψ
n
 → 0 (6.53)
since (A−λ)ψ
n
 → 0 by assumption and Kψ
n
 → 0 by Lemma 6.8 (iii).
Hence σ
ess
(A) ⊆ σ
ess
(A + K). Reversing the roles of A + K and A shows
σ
ess
(A+K) = σ
ess
(A). Since we have shown that we can remove any point
in the discrete spectrum by a selfadjoint ﬁnite rank operator we obtain the
following equivalent characterization of the essential spectrum.
Lemma 6.17. The essential spectrum of a selfadjoint operator A is pre
cisely the part which is invariant under rankone perturbations. In particu
lar,
σ
ess
(A) =
¸
K∈C(H),K
∗
=K
σ(A+K). (6.54)
There is even a larger class of operators under which the essential spec
trum is invariant.
Theorem 6.18 (Weyl). Suppose A and B are selfadjoint operators. If
R
A
(z) −R
B
(z) ∈ C(H) (6.55)
for one z ∈ ρ(A) ∩ ρ(B), then
σ
ess
(A) = σ
ess
(B). (6.56)
Proof. In fact, suppose λ ∈ σ
ess
(A) and let ψ
n
be a corresponding singular
Weyl sequence. Then (R
A
(z)−
1
λ−z
)ψ
n
=
R
A
(z)
z−λ
(A−λ)ψ
n
and thus (R
A
(z)−
1
λ−z
)ψ
n
 → 0. Moreover, by our assumption we also have (R
B
(z) −
1
λ−z
)ψ
n
 → 0 and thus (B − λ)ϕ
n
 → 0, where ϕ
n
= R
B
(z)ψ
n
. Since
130 6. Perturbation theory for selfadjoint operators
lim
n→∞
ϕ
n
 = lim
n→∞
R
A
(z)ψ
n
 = [λ − z[
−1
= 0 (since (R
A
(z) −
1
λ−z
)ψ
n
 = 
1
λ−z
R
A
(z)(A−λ)ψ
n
 → 0) we obtain a singular Weyl sequence
for B, showing λ ∈ σ
ess
(B). Now interchange the roles of A and B.
As a ﬁrst consequence note the following result
Theorem 6.19. Suppose A is symmetric with equal ﬁnite defect indices,
then all selfadjoint extensions have the same essential spectrum.
Proof. By Lemma 2.27 the resolvent diﬀerence of two selfadjoint extensions
is a ﬁnite rank operator if the defect indices are ﬁnite.
In addition, the following result is of interest.
Lemma 6.20. Suppose
R
A
(z) −R
B
(z) ∈ C(H) (6.57)
for one z ∈ ρ(A)∩ρ(B), then this holds for all z ∈ ρ(A)∩ρ(B). In addition,
if A and B are selfadjoint, then
f(A) −f(B) ∈ C(H) (6.58)
for any f ∈ C
∞
(R).
Proof. If the condition holds for one z it holds for all since we have (using
both resolvent formulas)
R
A
(z
) −R
B
(z
)
= (1 −(z −z
)R
B
(z
))(R
A
(z) −R
B
(z))(1 −(z −z
)R
A
(z
)). (6.59)
Let A and B be selfadjoint. The set of all functions f for which the
claim holds is a closed ∗subalgebra of C
∞
(R) (with sup norm). Hence the
claim follows from Lemma 4.4.
Remember that we have called K relatively compact with respect to A
if KR
A
(z) is compact (for one and hence for all z) and note that the the
resolvent diﬀerence R
A+K
(z) −R
A
(z) is compact if K is relatively compact.
In particular, Theorem 6.18 applies if B = A + K, where K is relatively
compact.
For later use observe that set of all operators which are relatively com
pact with respect to A forms a linear space (since compact operators do)
and relatively compact operators have Abound zero.
Lemma 6.21. Let A be selfadjoint and suppose K is relatively compact
with respect to A. Then the Abound of K is zero.
6.5. Strong and norm resolvent convergence 131
Proof. Write
KR
A
(λi) = (KR
A
(i))((A+ i)R
A
(λi)) (6.60)
and observe that the ﬁrst operator is compact and the second is normal
and converges strongly to 0 (by the spectral theorem). Hence the claim
follows from Lemma 6.3 and the discussion after Lemma 6.8 (since R
A
is
normal).
In addition, note the following result which is a straightforward conse
quence of the second resolvent identity.
Lemma 6.22. Suppose A is selfadjoint and B is symmetric with Abound
less then one. If K is relatively compact with respect to A then it is also
relatively compact with respect to A+B.
Proof. Since B is A bounded with Abound less than one, we can choose a
z ∈ C such that BR
A
(z) < 1. And hence
BR
A+B
(z) = BR
A
(z)(I +BR
A
(z))
−1
(6.61)
shows that B is also A+B bounded and the result follows from
KR
A+B
(z) = KR
A
(z)(I −BR
A+B
(z)) (6.62)
since KR
A
(z) is compact and BR
A+B
(z) is bounded.
Problem 6.8. Show that A = −
d
2
dx
2
+q(x), D(A) = H
2
(R) is selfadjoint if
q ∈ L
∞
(R). Show that if −u
(x)+q(x)u(x) = zu(x) has a solution for which
u and u
are bounded near +∞ (or −∞) but u is not square integrable near
+∞ (or −∞), then z ∈ σ
ess
(A). (Hint: Use u to construct a Weyl sequence
by restricting it to a compact set. Now modify your construction to get a
singular Weyl sequence by observing that functions with disjoint support are
orthogonal.)
6.5. Strong and norm resolvent convergence
Suppose A
n
and A are selfadjoint operators. We say that A
n
converges to
A in norm respectively strong resolvent sense if
lim
n→∞
R
An
(z) = R
A
(z) respectively slim
n→∞
R
An
(z) = R
A
(z) (6.63)
for one z ∈ Γ = C`Σ, Σ = σ(A) ∪
¸
n
σ(A
n
).
Using the Stone–Weierstraß theorem we obtain as a ﬁrst consequence
Theorem 6.23. Suppose A
n
converges to A in norm resolvent sense, then
f(A
n
) converges to f(A) in norm for any bounded continuous function
f : Σ → C with lim
λ→−∞
f(λ) = lim
λ→∞
f(λ). If A
n
converges to A
in strong resolvent sense, then f(A
n
) converges to f(A) strongly for any
bounded continuous function f : Σ →C.
132 6. Perturbation theory for selfadjoint operators
Proof. The set of functions for which the claim holds clearly forms a ∗
algebra (since resolvents are normal, taking adjoints is continuous even with
respect to strong convergence) and since it contains f(λ) = 1 and f(λ) =
1
λ−z
0
this ∗algebra is dense by the Stone–Weierstaß theorem. The usual
ε
3
shows that this ∗algebra is also closed.
To see the last claim let χ
n
be a compactly supported continuous func
tion (0 ≤ χ
m
≤ 1) which is one on the interval [−m, m]. Then f(A
n
)χ
m
(A
n
)
s
→
f(A)χ
m
(A) by the ﬁrst part and hence
(f(A
n
) −f(A))ψ ≤ f(A
n
) (1 −χ
m
(A
n
))ψ
+f(A
n
) (χ
m
(A
n
) −χ
m
(A))ψ
+(f(A
n
)χ
m
(A
n
) −f(A)χ
m
(A))ψ
+f(A) (1 −χ
m
(A))ψ (6.64)
can be made arbitrarily small since f(.) ≤ f
∞
and χ
m
(.)
s
→I by Theo
rem 3.1.
As a consequence note that the point z ∈ Γ is of no importance
Corollary 6.24. Suppose A
n
converges to A in norm or strong resolvent
sense for one z
0
∈ Γ, then this holds for all z ∈ Γ.
and that we have
Corollary 6.25. Suppose A
n
converges to A in strong resolvent sense, then
e
itAn
s
→ e
itA
, t ∈ R, (6.65)
and if all operators are semibounded
e
−tAn
s
→ e
−tA
, t ≥ 0. (6.66)
Finally we need some good criteria to check for norm respectively strong
resolvent convergence.
Lemma 6.26. Let A
n
, A be selfadjoint operators with D(A
n
) = D(A).
Then A
n
converges to A in norm resolvent sense if there are sequences a
n
and b
n
converging to zero such that
(A
n
−A)ψ ≤ a
n
ψ +b
n
Aψ, ψ ∈ D(A) = D(A
n
). (6.67)
Proof. From the second resolvent identity
R
An
(z) −R
A
(z) = R
An
(z)(A−A
n
)R
A
(z) (6.68)
we infer
(R
An
(i) −R
A
(i))ψ ≤ R
An
(i)
a
n
R
A
(i)ψ +b
n
AR
A
(i)ψ
≤ (a
n
+b
n
)ψ (6.69)
6.5. Strong and norm resolvent convergence 133
and hence R
An
(i) −R
A
(i) ≤ a
n
+b
n
→ 0.
In particular, norm convergence implies norm resolvent convergence:
Corollary 6.27. Let A
n
, A be bounded selfadjoint operators with A
n
→ A,
then A
n
converges to A in norm resolvent sense.
Similarly, if no domain problems get in the way, strong convergence
implies strong resolvent convergence:
Lemma 6.28. Let A
n
, A be selfadjoint operators. Then A
n
converges to
A in strong resolvent sense if there there is a core D
0
of A such that for any
ψ ∈ D
0
we have ψ ∈ D(A
n
) for n suﬃciently large and A
n
ψ → Aψ.
Proof. Using the second resolvent identity we have
(R
An
(i) −R
A
(i))ψ ≤ (A−A
n
)R
A
(i)ψ → 0 (6.70)
for ψ ∈ (A −i)D
0
which is dense, since D
0
is a core. The rest follows from
Lemma 1.13.
If you wonder why we did not deﬁne weak resolvent convergence, here
is the answer: it is equivalent to strong resolvent convergence.
Lemma 6.29. Suppose wlim
n→∞
R
An
(z) = R
A
(z) for some z ∈ Γ, then
also slim
n→∞
R
An
(z) = R
A
(z).
Proof. By R
An
(z) R
A
(z) we have also R
An
(z)
∗
R
A
(z)
∗
and thus by
the ﬁrst resolvent identity
R
An
(z)ψ
2
−R
A
(z)ψ
2
= 'ψ, R
An
(z
∗
)R
An
(z)ψ −R
A
(z
∗
)R
A
(z)ψ`
=
1
z −z
∗
'ψ, (R
An
(z) −R
An
(z
∗
) +R
A
(z) −R
A
(z
∗
))ψ` → 0. (6.71)
Together with R
An
(z)ψ R
A
(z)ψ we have R
An
(z)ψ → R
A
(z)ψ by virtue
of Lemma 1.11 (iv).
Now what can we say about the spectrum?
Theorem 6.30. Let A
n
and A be selfadjoint operators. If A
n
converges to
A in strong resolvent sense we have σ(A) ⊆ lim
n→∞
σ(A
n
). If A
n
converges
to A in norm resolvent sense we have σ(A) = lim
n→∞
σ(A
n
).
Proof. Suppose the ﬁrst claim were wrong. Then we can ﬁnd a λ ∈ σ(A)
and some ε > 0 such that σ(A
n
) ∩ (λ − ε, λ + ε) = ∅. Choose a bounded
continuous function f which is one on (λ −
ε
2
, λ +
ε
2
) and vanishes outside
(λ−ε, λ+ε). Then f(A
n
) = 0 and hence f(A)ψ = limf(A
n
)ψ = 0 for every
ψ. On the other hand, since λ ∈ σ(A) there is a nonzero ψ ∈ RanP
A
((λ −
ε
2
, λ +
ε
2
)) implying f(A)ψ = ψ, a contradiction.
134 6. Perturbation theory for selfadjoint operators
To see the second claim, recall that the norm of R
A
(z) is just one over
the distance from the spectrum. In particular, λ ∈ σ(A) if and only if
R
A
(λ + i) < 1. So λ ∈ σ(A) implies R
A
(λ + i) < 1, which implies
R
An
(λ + i) < 1 for n suﬃciently large, which implies λ ∈ σ(A
n
) for n
suﬃciently large.
Note that the spectrum can contract if we only have strong resolvent
sense: Let A
n
be multiplication by
1
n
x in L
2
(R). Then A
n
converges to 0 in
strong resolvent sense, but σ(A
n
) = R and σ(0) = ¦0¦.
Lemma 6.31. Suppose A
n
converges in strong resolvent sense to A. If
P
A
(¦λ¦) = 0, then
slim
n→∞
P
An
((−∞, λ)) = slim
n→∞
P
An
((−∞, λ]) = P
A
((−∞, λ)) = P
A
((−∞, λ]).
(6.72)
Proof. The idea is to approximate χ
(−∞,λ)
by a continuous function f with
0 ≤ f ≤ χ
(−∞,λ)
. Then
(P
A
((−∞, λ)) −P
An
((−∞, λ)))ψ ≤ (P
A
((−∞, λ)) −f(A))ψ
+(f(A) −f(A
n
))ψ +(f(A
n
) −P
An
((−∞, λ)))ψ. (6.73)
The ﬁrst term can be made arbitrarily small if we let f converge pointwise
to χ
(−∞,λ)
(Theorem 3.1) and the same is true for the second if we choose n
large (Theorem 6.23). However, the third term can only be made small for
ﬁxed n. To overcome this problem let us choose another continuous function
g with χ
(−∞,λ]
≤ g ≤ 1. Then
(f(A
n
) −P
An
((−∞, λ)))ψ ≤ (g(A
n
) −f(A
n
))ψ (6.74)
since f ≤ χ
(−∞,λ)
≤ χ
(−∞,λ]
≤ g. Furthermore,
(g(A
n
) −f(A
n
))ψ ≤ (g(A
n
) −f(A))ψ
+(f(A) −g(A))ψ +(g(A) −g(A
n
))ψ (6.75)
and now all terms are under control. Since we can replace P
.
((−∞, λ)) by
P
.
((−∞, λ]) in all calculations we are done.
Using P((λ
0
, λ
1
)) = P((−∞, λ
1
)) −P((−∞, λ
0
]) we also obtain
Corollary 6.32. Suppose A
n
converges in strong resolvent sense to A. If
P
A
(¦λ
0
¦) = P
A
(¦λ
1
¦) = 0, then
slim
n→∞
P
An
((λ
0
, λ
1
)) = slim
n→∞
P
An
([λ
0
, λ
1
]) = P
A
((λ
0
, λ
1
)) = P
A
([λ
0
, λ
1
]).
(6.76)
6.5. Strong and norm resolvent convergence 135
Example. The following example shows that the requirement P
A
(¦λ¦) = 0
is crucial, even if we have bounded operators and norm convergence. In fact,
let H = C
2
and
A
n
=
1
n
1 0
0 −1
. (6.77)
Then A
n
→ 0 and
P
An
((−∞, 0)) = P
An
((−∞, 0]) =
0 0
0 1
, (6.78)
but P
0
((−∞, 0)) = 0 and P
0
((−∞, 0]) = I.
Problem 6.9. Show that for selfadjoint operators, strong resolvent conver
gence is equivalent to convergence with respect to the metric
d(A, B) =
¸
n∈N
1
2
n
(R
A
(i) −R
B
(i))ϕ
n
, (6.79)
where ¦ϕ
n
¦
n∈N
is some (ﬁxed) ONB.
Problem 6.10 (Weak convergence of spectral measures). Suppose A
n
→
A in strong resolvent sense and let µ
ψ
, µ
ψ,n
be the corresponding spectral
measures. Show that
f(λ)dµ
ψ,n
(λ) →
f(λ)dµ
ψ
(λ) (6.80)
for every bounded continuous f. Give a counterexample when f is not con
tinuous.
Part 2
Schr¨odinger Operators
Chapter 7
The free Schr¨odinger
operator
7.1. The Fourier transform
We ﬁrst review some basic facts concerning the Fourier transform which
will be needed in the following section.
Let C
∞
(R
n
) be the set of all complexvalued functions which have partial
derivatives of arbitrary order. For f ∈ C
∞
(R
n
) and α ∈ N
n
0
we set
∂
α
f =
∂
α
f
∂x
α
1
1
∂x
αn
n
, x
α
= x
α
1
1
x
αn
n
, [α[ = α
1
+ +α
n
. (7.1)
An element α ∈ N
n
0
is called multiindex and [α[ is called its order. Recall
the Schwarz space
o(R
n
) = ¦f ∈ C
∞
(R
n
)[ sup
x
[x
α
(∂
β
f)(x)[ < ∞, α, β ∈ N
n
0
¦ (7.2)
which is dense in L
2
(R
n
) (since C
∞
c
(R
n
) ⊂ o(R
n
) is). For f ∈ o(R
n
) we
deﬁne
T(f)(p) ≡
ˆ
f(p) =
1
(2π)
n/2
R
n
e
−ipx
f(x)d
n
x. (7.3)
Then it is an exercise in partial integration to prove
Lemma 7.1. For any multiindex α ∈ N
n
0
and any f ∈ o(R
n
) we have
(∂
α
f)
∧
(p) = (ip)
α
ˆ
f(p), (x
α
f(x))
∧
(p) = i
α
∂
α
ˆ
f(p). (7.4)
Hence we will sometimes write pf(x) for −i∂f(x), where ∂ = (∂
1
, . . . , ∂
n
)
is the gradient.
139
140 7. The free Schr¨odinger operator
In particular T maps o(R
n
) into itself. Another useful property is the
convolution formula.
Lemma 7.2. The Fourier transform of the convolution
(f ∗ g)(x) =
R
n
f(y)g(x −y)d
n
y =
R
n
f(x −y)g(y)d
n
y (7.5)
of two functions f, g ∈ o(R
n
) is given by
(f ∗ g)
∧
(p) = (2π)
n/2
ˆ
f(p)ˆ g(p). (7.6)
Proof. We compute
(f ∗ g)
∧
(p) =
1
(2π)
n/2
R
n
e
−ipx
R
n
f(y)g(x −y)d
n
y d
n
x
=
R
n
e
−ipy
f(y)
1
(2π)
n/2
R
n
e
−ip(x−y)
g(x −y)d
n
xd
n
y
=
R
n
e
−ipy
f(y)ˆ g(p)d
n
y = (2π)
n/2
ˆ
f(p)ˆ g(p), (7.7)
where we have used Fubini’s theorem.
Next, we want to compute the inverse of the Fourier transform. For this
the following lemma will be needed.
Lemma 7.3. We have e
−zx
2
/2
∈ o(R
n
) for Re(z) > 0 and
T(e
−zx
2
/2
)(p) =
1
z
n/2
e
−p
2
/(2z)
. (7.8)
Here z
n/2
has to be understood as (
√
z)
n
, where the branch cut of the root
is chosen along the negative real axis.
Proof. Due to the product structure of the exponential, one can treat each
coordinate separately, reducing the problem to the case n = 1.
Let φ
z
(x) = exp(−zx
2
/2). Then φ
z
(x)+zxφ
z
(x) = 0 and hence i(p
ˆ
φ
z
(p)+
z
ˆ
φ
z
(p)) = 0. Thus
ˆ
φ
z
(p) = cφ
1/z
(p) and (Problem 7.1)
c =
ˆ
φ
z
(0) =
1
√
2π
R
exp(−zx
2
/2)dx =
1
√
z
(7.9)
at least for z > 0. However, since the integral is holomorphic for Re(z) > 0,
this holds for all z with Re(z) > 0 if we choose the branch cut of the root
along the negative real axis.
Now we can show
7.1. The Fourier transform 141
Theorem 7.4. The Fourier transform T : o(R
n
) → o(R
n
) is a bijection.
Its inverse is given by
T
−1
(g)(x) ≡ ˇ g(x) =
1
(2π)
n/2
R
n
e
ipx
g(p)d
n
p. (7.10)
We have T
2
(f)(x) = f(−x) and thus T
4
= I.
Proof. It suﬃces to show T
2
(f)(x) = f(−x). Consider φ
z
(x) from the
proof of the previous lemma and observe T
2
(φ
z
)(x) = φ
z
(−x). Moreover,
using Fubini this even implies T
2
(f
ε
)(x) = f
ε
(−x) for any ε > 0, where
f
ε
(x) =
1
ε
n
R
n
φ
1/ε
2(x −y)f(y)d
n
y. (7.11)
Since lim
ε↓0
f
ε
(x) = f(x) for every x ∈ R
n
(Problem 7.2), we infer from
dominated convergence T
2
(f)(x) = lim
ε↓0
T
2
(f
ε
)(x) = lim
ε↓0
f
ε
(−x) =
f(−x).
From Fubini’s theorem we also obtain Parseval’s identity
R
n
[
ˆ
f(p)[
2
d
n
p =
1
(2π)
n/2
R
n
R
n
f(x)
∗
ˆ
f(p)e
ipx
d
n
p d
n
x
=
R
n
[f(x)[
2
d
n
x (7.12)
for f ∈ o(R
n
) and thus we can extend T to a unitary operator:
Theorem 7.5. The Fourier transform T extends to a unitary operator T :
L
2
(R
n
) → L
2
(R
n
). Its spectrum satisﬁes
σ(T) = ¦z ∈ C[z
4
= 1¦ = ¦1, −1, i, −i¦. (7.13)
Proof. It remains to compute the spectrum. In fact, if ψ
n
is a Weyl se
quence, then (T
2
+ z
2
)(T + z)(T − z)ψ
n
= (T
4
− z
4
)ψ
n
= (1 − z
4
)ψ
n
→ 0
implies z
4
= 1. Hence σ(T) ⊆ ¦z ∈ C[z
4
= 1¦. We defer the proof for
equality to Section 8.3, where we will explicitly compute an orthonormal
basis of eigenfunctions.
Lemma 7.1 also allows us to extend diﬀerentiation to a larger class. Let
us introduce the Sobolev space
H
r
(R
n
) = ¦f ∈ L
2
(R
n
)[[p[
r
ˆ
f(p) ∈ L
2
(R
n
)¦. (7.14)
We will abbreviate
∂
α
f = ((ip)
α
ˆ
f(p))
∨
, f ∈ H
r
(R
n
), [α[ ≤ r (7.15)
which implies
R
n
g(x)(∂
α
f)(x)d
n
x = (−1)
α
R
n
(∂
α
g)(x)f(x)d
n
x, (7.16)
142 7. The free Schr¨odinger operator
for f ∈ H
r
(R
n
) and g ∈ o(R
n
). That is, ∂
α
f is the derivative of f in the
sense of distributions.
Finally, we have the RiemannLebesgue lemma.
Lemma 7.6 (RiemannLebesgue). Let C
∞
(R
n
) denote the Banach space of
all continuous functions f : R
n
→ C which vanish at ∞ equipped with the
sup norm. Then the Fourier transform is a bounded map from L
1
(R
n
) into
C
∞
(R
n
) satisfying

ˆ
f
∞
≤ (2π)
−n/2
f
1
. (7.17)
Proof. Clearly we have
ˆ
f ∈ C
∞
(R
n
) if f ∈ o(R
n
). Moreover, the estimate
sup
p
[
ˆ
f(p)[ ≤
1
(2π)
n/2
sup
p
R
n
[e
−ipx
f(x)[d
n
x =
1
(2π)
n/2
R
n
[f(x)[d
n
x
(7.18)
shows
ˆ
f ∈ C
∞
(R
n
) for arbitrary f ∈ L
1
(R
n
) since o(R
n
) is dense in L
1
(R
n
).
Problem 7.1. Show that
R
exp(−x
2
/2)dx =
√
2π. (Hint: Square the inte
gral and evaluate it using polar coordinates.)
Problem 7.2. Extend Lemma 0.32 to the case u, f ∈ o(R
n
) (compare Prob
lem 0.17).
7.2. The free Schr¨odinger operator
In Section 2.1 we have seen that the Hilbert space corresponding to one
particle in R
3
is L
2
(R
3
). More generally, the Hilbert space for N particles
in R
d
is L
2
(R
n
), n = Nd. The corresponding non relativistic Hamilton
operator, if the particles do not interact, is given by
H
0
= −∆, (7.19)
where ∆ is the Laplace operator
∆ =
n
¸
j=1
∂
2
∂x
2
j
. (7.20)
Our ﬁrst task is to ﬁnd a good domain such that H
0
is a selfadjoint operator.
By Lemma 7.1 we have that
−∆ψ(x) = (p
2
ˆ
ψ(p))
∨
(x), ψ ∈ H
2
(R
n
), (7.21)
and hence the operator
H
0
ψ = −∆ψ, D(H
0
) = H
2
(R
n
), (7.22)
7.2. The free Schr¨odinger operator 143
is unitarily equivalent to the maximally deﬁned multiplication operator
(T H
0
T
−1
)ϕ(p) = p
2
ϕ(p), D(p
2
) = ¦ϕ ∈ L
2
(R
n
)[p
2
ϕ(p) ∈ L
2
(R
n
)¦.
(7.23)
Theorem 7.7. The free Schr¨odinger operator H
0
is selfadjoint and its
spectrum is characterized by
σ(H
0
) = σ
ac
(H
0
) = [0, ∞), σ
sc
(H
0
) = σ
pp
(H
0
) = ∅. (7.24)
Proof. It suﬃces to show that dµ
ψ
is purely absolutely continuous for every
ψ. First observe that
'ψ, R
H
0
(z)ψ` = '
ˆ
ψ, R
p
2(z)
ˆ
ψ` =
R
n
[
ˆ
ψ(p)[
2
p
2
−z
d
n
p =
R
1
r
2
−z
d˜ µ
ψ
(r), (7.25)
where
d˜ µ
ψ
(r) = χ
[0,∞)
(r)r
n−1
S
n−1
[
ˆ
ψ(rω)[
2
d
n−1
ω
dr. (7.26)
Hence, after a change of coordinates, we have
'ψ, R
H
0
(z)ψ` =
R
1
λ −z
dµ
ψ
(λ), (7.27)
where
dµ
ψ
(λ) =
1
2
χ
[0,∞)
(λ)λ
n/2−1
S
n−1
[
ˆ
ψ(
√
λω)[
2
d
n−1
ω
dλ, (7.28)
proving the claim.
Finally, we note that the compactly supported smooth functions are a
core for H
0
.
Lemma 7.8. The set C
∞
c
(R
n
) = ¦f ∈ o(R
n
)[supp(f) is compact¦ is a core
for H
0
.
Proof. It is not hard to see that o(R
n
) is a core (Problem 7.3) and hence it
suﬃces to show that the closure of H
0
[
C
∞
c
(R
n
)
contains H
0
[
S(R
n
)
. To see this,
let ϕ(x) ∈ C
∞
c
(R
n
) which is one for [x[ ≤ 1 and vanishes for [x[ ≥ 2. Set
ϕ
n
(x) = ϕ(
1
n
x), then ψ
n
(x) = ϕ
n
(x)ψ(x) is in C
∞
c
(R
n
) for every ψ ∈ o(R
n
)
and ψ
n
→ ψ respectively ∆ψ
n
→ ∆ψ.
Note also that the quadratic form of H
0
is given by
q
H
0
(ψ) =
n
¸
j=1
R
n
[∂
j
ψ(x)[
2
d
n
x, ψ ∈ Q(H
0
) = H
1
(R
n
). (7.29)
Problem 7.3. Show that o(R
n
) is a core for H
0
. (Hint: Show that the
closure of H
0
[
S(R
n
)
contains H
0
.)
144 7. The free Schr¨odinger operator
Problem 7.4. Show that ¦ψ ∈ o(R)[ψ(0) = 0¦ is dense but not a core for
H
0
= −
d
2
dx
2
.
7.3. The time evolution in the free case
Now let us look at the time evolution. We have
e
−itH
0
ψ(x) = T
−1
e
−itp
2
ˆ
ψ(p). (7.30)
The right hand side is a product and hence our operator should be express
ible as an integral operator via the convolution formula. However, since
e
−itp
2
is not in L
2
, a more careful analysis is needed.
Consider
f
ε
(p
2
) = e
−(it+ε)p
2
, ε > 0. (7.31)
Then f
ε
(H
0
)ψ → e
−itH
0
ψ by Theorem 3.1. Moreover, by Lemma 7.3 and
the convolution formula we have
f
ε
(H
0
)ψ(x) =
1
(4π(it +ε))
n/2
R
n
e
−
x−y
2
4(it+ε)
ψ(y)d
n
y (7.32)
and hence
e
−itH
0
ψ(x) =
1
(4πit)
n/2
R
n
e
i
x−y
2
4t
ψ(y)d
n
y (7.33)
for t = 0 and ψ ∈ L
1
∩ L
2
. For general ψ ∈ L
2
the integral has to be
understood as a limit.
Using this explicit form, it is not hard to draw some immediate conse
quences. For example, if ψ ∈ L
2
(R
n
) ∩L
1
(R
n
), then ψ(t) ∈ C(R
n
) for t = 0
(use dominated convergence and continuity of the exponential) and satisﬁes
ψ(t)
∞
≤
1
[4πt[
n/2
ψ(0)
1
(7.34)
by the RiemannLebesgue lemma. Thus we have spreading of wave functions
in this case. Moreover, it is even possible to determine the asymptotic form
of the wave function for large t as follows. Observe
e
−itH
0
ψ(x) =
e
i
x
2
4t
(4πit)
n/2
R
n
e
i
y
2
4t
ψ(y)e
i
xy
2t
d
n
y
=
1
2it
n/2
e
i
x
2
4t
e
i
y
2
4t
ψ(y)
∧
(
x
2t
). (7.35)
Moreover, since exp(i
y
2
4t
)ψ(y) → ψ(y) in L
2
as [t[ → ∞ (dominated conver
gence) we obtain
7.4. The resolvent and Green’s function 145
Lemma 7.9. For any ψ ∈ L
2
(R
n
) we have
e
−itH
0
ψ(x) −
1
2it
n/2
e
i
x
2
4t
ˆ
ψ(
x
2t
) → 0 (7.36)
in L
2
as [t[ → ∞.
Next we want to apply the RAGE theorem in order to show that for any
initial condition, a particle will escape to inﬁnity.
Lemma 7.10. Let g(x) be the multiplication operator by g and let f(p) be
the operator given by f(p)ψ(x) = T
−1
(f(p)
ˆ
ψ(p))(x). Denote by L
∞
∞
(R
n
) the
bounded Borel functions which vanish at inﬁnity. Then
f(p)g(x) and g(x)f(p) (7.37)
are compact if f, g ∈ L
∞
∞
(R
n
) and (extend to) Hilbert–Schmidt operators if
f, g ∈ L
2
(R
n
).
Proof. By symmetry it suﬃces to consider g(x)f(p). Let f, g ∈ L
2
, then
g(x)f(p)ψ(x) =
1
(2π)
n/2
R
n
g(x)
ˇ
f(x −y)ψ(y)d
n
y (7.38)
shows that g(x)f(p) is Hilbert–Schmidt since g(x)
ˇ
f(x −y) ∈ L
2
(R
n
R
n
).
If f, g are bounded then the functions f
R
(p) = χ
{pp
2
≤R}
(p)f(p) and
g
R
(x) = χ
{xx
2
≤R}
(x)g(x) are in L
2
. Thus g
R
(x)f
R
(p) is compact and tends
to g(x)f(p) in norm since f, g vanish at inﬁnity.
In particular, this lemma implies that
χ
Ω
(H
0
+ i)
−1
(7.39)
is compact if Ω ⊆ R
n
is bounded and hence
lim
t→∞
χ
Ω
e
−itH
0
ψ
2
= 0 (7.40)
for any ψ ∈ L
2
(R
n
) and any bounded subset Ω of R
n
. In other words, the
particle will eventually escape to inﬁnity since the probability of ﬁnding the
particle in any bounded set tends to zero. (If ψ ∈ L
1
(R
n
) this of course also
follows from (7.34).)
7.4. The resolvent and Green’s function
Now let us compute the resolvent of H
0
. We will try to use a similar approach
as for the time evolution in the previous section. However, since it is highly
nontrivial to compute the inverse Fourier transform of exp(−εp
2
)(p
2
−z)
−1
directly, we will use a small ruse.
146 7. The free Schr¨odinger operator
Note that
R
H
0
(z) =
∞
0
e
zt
e
−tH
0
dt, Re(z) < 0 (7.41)
by Lemma 4.1. Moreover,
e
−tH
0
ψ(x) =
1
(4πt)
n/2
R
n
e
−
x−y
2
4t
ψ(y)d
n
y, t > 0, (7.42)
by the same analysis as in the previous section. Hence, by Fubini, we have
R
H
0
(z)ψ(x) =
R
n
G
0
(z, [x −y[)ψ(y)d
n
y, (7.43)
where
G
0
(z, r) =
∞
0
1
(4πt)
n/2
e
−
r
2
4t
+zt
dt, r > 0, Re(z) < 0. (7.44)
The function G
0
(z, r) is called Green’s function of H
0
. The integral can
be evaluated in terms of modiﬁed Bessel functions of the second kind
G
0
(z, r) =
1
2π
−z
4π
2
r
2
n−2
4
K
n
2
−1
(
√
−zr). (7.45)
The functions K
ν
(x) satisfy the following diﬀerential equation
d
2
dx
2
+
1
x
d
dx
−1 −
ν
2
x
2
K
ν
(x) = 0 (7.46)
and have the following asymptotics
K
ν
(x) =
Γ(ν)
2
x
2
−ν
+O(x
−ν+1
) ν = 0
−ln(
x
2
) +O(1) ν = 0
(7.47)
for [x[ → 0 and
K
ν
(x) =
π
2x
e
−x
(1 +O(x
−1
)) (7.48)
for [x[ → ∞. For more information see for example [22]. In particular,
G
0
(z, r) has an analytic continuation for z ∈ C`[0, ∞) = ρ(H
0
). Hence we
can deﬁne the right hand side of (7.43) for all z ∈ ρ(H
0
) such that
R
n
R
n
ϕ(x)G
0
(z, [x −y[)ψ(y)d
n
yd
n
x (7.49)
is analytic for z ∈ ρ(H
0
) and ϕ, ψ ∈ o(R
n
) (by Morea’s theorem). Since
it is equal to 'ϕ, R
H
0
(z)ψ` for Re(z) < 0 it is equal to this function for all
z ∈ ρ(H
0
), since both functions are analytic in this domain. In particular,
(7.43) holds for all z ∈ ρ(H
0
).
7.4. The resolvent and Green’s function 147
If n is odd, we have the case of spherical Bessel functions which can be
expressed in terms of elementary functions. For example, we have
G
0
(z, r) =
1
2
√
−z
e
−
√
−z r
, n = 1, (7.50)
and
G
0
(z, r) =
1
4πr
e
−
√
−z r
, n = 3. (7.51)
Problem 7.5. Verify (7.43) directly in the case n = 1.
Chapter 8
Algebraic methods
8.1. Position and momentum
Apart from the Hamiltonian H
0
, which corresponds to the kinetic energy,
there are several other important observables associated with a single parti
cle in three dimensions. Using commutation relation between these observ
ables, many important consequences about these observables can be derived.
First consider the oneparameter unitary group
(U
j
(t)ψ)(x) = e
−itx
j
ψ(x), 1 ≤ j ≤ 3. (8.1)
For ψ ∈ o(R
3
) we compute
lim
t→0
i
e
−itx
j
ψ(x) −ψ(x)
t
= x
j
ψ(x) (8.2)
and hence the generator is the multiplication operator by the jth coordinate
function. By Corollary 5.3 it is essentially selfadjoint on ψ ∈ o(R
3
). It is
custom to combine all three operators to one vector valued operator x, which
is known as position operator. Moreover, it is not hard to see that the
spectrum of x
j
is purely absolutely continuous and given by σ(x
j
) = R. In
fact, let ϕ(x) be an orthonormal basis for L
2
(R). Then ϕ
i
(x
1
)ϕ
j
(x
2
)ϕ
k
(x
3
)
is an orthonormal basis for L
2
(R
3
) and x
1
can be written as a orthogonal
sum of operators restricted to the subspaces spanned by ϕ
j
(x
2
)ϕ
k
(x
3
). Each
subspace is unitarily equivalent to L
2
(R) and x
1
is given by multiplication
with the identity. Hence the claim follows (or use Theorem 4.12).
Next, consider the oneparameter unitary group of translations
(U
j
(t)ψ)(x) = ψ(x −te
j
), 1 ≤ j ≤ 3, (8.3)
149
150 8. Algebraic methods
where e
j
is the unit vector in the jth coordinate direction. For ψ ∈ o(R
3
)
we compute
lim
t→0
i
ψ(x −te
j
) −ψ(x)
t
=
1
i
∂
∂x
j
ψ(x) (8.4)
and hence the generator is p
j
=
1
i
∂
∂x
j
. Again it is essentially selfadjoint
on ψ ∈ o(R
3
). Moreover, since it is unitarily equivalent to x
j
by virtue of
the Fourier transform we conclude that the spectrum of p
j
is again purely
absolutely continuous and given by σ(p
j
) = R. The operator p is known as
momentum operator. Note that since
[H
0
, p
j
]ψ(x) = 0, ψ ∈ o(R
3
) (8.5)
we have
d
dt
'ψ(t), p
j
ψ(t)` = 0, ψ(t) = e
−itH
0
ψ(0) ∈ o(R
3
), (8.6)
that is, the momentum is a conserved quantity for the free motion. Similarly
one has
[p
j
, x
k
]ψ(x) = δ
jk
ψ(x), ψ ∈ o(R
3
), (8.7)
which is known as the Weyl relation.
The Weyl relations also imply that the meansquare deviation of position
and momentum cannot be made arbitrarily small simultaneously:
Theorem 8.1 (Heisenberg Uncertainty Principle). Suppose A and B are
two symmetric operators, then for any ψ ∈ D(AB) ∩ D(AB) we have
∆
ψ
(A)∆
ψ
(B) ≥
1
2
[E
ψ
([A, B])[ (8.8)
with equality if
(B −E
ψ
(B))ψ = iλ(A−E
ψ
(A))ψ, λ ∈ R`¦0¦, (8.9)
or if ψ is an eigenstate of A or B.
Proof. Let us ﬁx ψ ∈ D(AB) ∩ D(AB) and abbreviate
ˆ
A = A−E
ψ
(A),
ˆ
B = B −E
ψ
(B). (8.10)
Then ∆
ψ
(A) = 
ˆ
Aψ, ∆
ψ
(B) = 
ˆ
Bψ and hence by CauchySchwarz
['
ˆ
Aψ,
ˆ
Bψ`[ ≤ ∆
ψ
(A)∆
ψ
(B). (8.11)
Now note that
ˆ
A
ˆ
B =
1
2
¦
ˆ
A,
ˆ
B¦ +
1
2
[A, B], ¦
ˆ
A,
ˆ
B¦ =
ˆ
A
ˆ
B +
ˆ
B
ˆ
A (8.12)
where ¦
ˆ
A,
ˆ
B¦ and i[A, B] are symmetric. So
['
ˆ
Aψ,
ˆ
Bψ`[
2
= ['ψ,
ˆ
A
ˆ
Bψ`[
2
=
1
2
['ψ, ¦
ˆ
A,
ˆ
B¦ψ`[
2
+
1
2
['ψ, [A, B]ψ`[
2
(8.13)
8.2. Angular momentum 151
which proves (8.8).
To have equality if ψ is not an eigenstate we need
ˆ
Bψ = z
ˆ
Aψ for equality
in CauchySchwarz and 'ψ, ¦
ˆ
A,
ˆ
B¦ψ` = 0. Inserting the ﬁrst into the second
requirement gives 0 = (z −z
∗
)
ˆ
Aψ
2
and shows Re(z) = 0.
In case of position and momentum we have (ψ = 1)
∆
ψ
(p
j
)∆
ψ
(x
k
) ≥
δ
jk
2
(8.14)
and the minimum is attained for the Gaussian wave packets
ψ(x) =
λ
π
n/4
e
−
λ
2
x−x
0

2
−ip
0
x
, (8.15)
which satisfy E
ψ
(x) = x
0
and E
ψ
(p) = p
0
respectively ∆
ψ
(p
j
)
2
=
λ
2
and
∆
ψ
(x
k
)
2
=
1
2λ
.
Problem 8.1. Check that (8.15) realizes the minimum.
8.2. Angular momentum
Now consider the oneparameter unitary group of rotations
(U
j
(t)ψ)(x) = ψ(M
j
(t)x), 1 ≤ j ≤ 3, (8.16)
where M
j
(t) is the matrix of rotation around e
j
by an angle of t. For
ψ ∈ o(R
3
) we compute
lim
t→0
i
ψ(M
i
(t)x) −ψ(x)
t
=
3
¸
j,k=1
ε
ijk
x
j
p
k
ψ(x), (8.17)
where
ε
ijk
=
1 if ijk is an even permutation of 123
−1 if ijk is an odd permutation of 123
0 else
. (8.18)
Again one combines the three components to one vector valued operator
L = x ∧ p, which is known as angular momentum operator. Since
e
i2πL
j
= I, we see that the spectrum is a subset of Z. In particular, the
continuous spectrum is empty. We will show below that we have σ(L
j
) = Z.
Note that since
[H
0
, L
j
]ψ(x) = 0, ψ ∈ o(R
3
), (8.19)
we have again
d
dt
'ψ(t), L
j
ψ(t)` = 0, ψ(t) = e
−itH
0
ψ(0) ∈ o(R
3
), (8.20)
that is, the angular momentum is a conserved quantity for the free motion
as well.
152 8. Algebraic methods
Moreover, we even have
[L
i
, K
j
]ψ(x) = i
3
¸
k=1
ε
ijk
K
k
ψ(x), ψ ∈ o(R
3
), K
j
∈ ¦L
j
, p
j
, x
j
¦, (8.21)
and these algebraic commutation relations are often used to derive informa
tion on the point spectra of these operators. In this respect the following
domain
D = span¦x
α
e
−
x
2
2
[ α ∈ N
n
0
¦ ⊂ o(R
n
) (8.22)
is often used. It has the nice property that the ﬁnite dimensional subspaces
D
k
= span¦x
α
e
−
x
2
2
[ [α[ ≤ k¦ (8.23)
are invariant under L
j
(and hence they reduce L
j
).
Lemma 8.2. The subspace D ⊂ L
2
(R
n
) deﬁned in (8.22) is dense.
Proof. By Lemma 1.9 it suﬃces to consider the case n = 1. Suppose
'ϕ, ψ` = 0 for every ψ ∈ D. Then
1
√
2π
ϕ(x)e
−
x
2
2
k
¸
j=1
(itx)
j
j!
= 0 (8.24)
for any ﬁnite k and hence also in the limit k → ∞ by the dominated conver
gence theorem. But the limit is the Fourier transform of ϕ(x)e
−
x
2
2
, which
shows that this function is zero. Hence ϕ(x) = 0.
Since it is invariant under the unitary groups generated by L
j
, the op
erators L
j
are essentially selfadjoint on D by Corollary 5.3.
Introducing L
2
= L
2
1
+L
2
2
+L
2
3
it is straightforward to check
[L
2
, L
j
]ψ(x) = 0, ψ ∈ o(R
3
). (8.25)
Moreover, D
k
is invariant under L
2
and L
3
and hence D
k
reduces L
2
and
L
3
. In particular, L
2
and L
3
are given by ﬁnite matrices on D
k
. Now
let H
m
= Ker(L
3
− m) and denote by P
k
the projector onto D
k
. Since
L
2
and L
3
commute on D
k
, the space P
k
H
m
is invariant under L
2
which
shows that we can choose an orthonormal basis consisting of eigenfunctions
of L
2
for P
k
H
m
. Increasing k we get an orthonormal set of simultaneous
eigenfunctions whose span is equal to D. Hence there is an orthonormal
basis of simultaneous eigenfunctions of L
2
and L
3
.
Now let us try to draw some further consequences by using the commuta
tion relations (8.21). (All commutation relations below hold for ψ ∈ o(R
3
).)
Denote by H
l,m
the set of all functions in D satisfying
L
3
ψ = mψ, L
2
ψ = l(l + 1)ψ. (8.26)
8.2. Angular momentum 153
By L
2
≥ 0 and σ(L
3
) ⊆ Z we can restrict our attention to the case l ≥ 0
and m ∈ Z.
First introduce two new operators
L
±
= L
1
±iL
2
, [L
3
, L
±
] = ±L
±
. (8.27)
Then, for every ψ ∈ H
l,m
we have
L
3
(L
±
ψ) = (m±1)(L
±
ψ), L
2
(L
±
ψ) = l(l + 1)(L
±
ψ), (8.28)
that is, L
±
H
l,m
→H
l,m±1
. Moreover, since
L
2
= L
2
3
±L
3
+L
∓
L
±
(8.29)
we obtain
L
±
ψ
2
= 'ψ, L
∓
L
±
ψ` = (l(l + 1) −m(m±1))ψ (8.30)
for every ψ ∈ H
l,m
. If ψ = 0 we must have l(l + 1) − m(m± 1) ≥ 0 which
shows H
l,m
= ¦0¦ for [m[ > l. Moreover, L
±
H
l,m
→ H
l,m±1
is injective
unless [m[ = l. Hence we must have H
l,m
= ¦0¦ for l ∈ N
0
.
Up to this point we know σ(L
2
) ⊆ ¦l(l +1)[l ∈ N
0
¦, σ(L
3
) ⊆ Z. In order
to show that equality holds in both cases, we need to show that H
l,m
= ¦0¦
for l ∈ N
0
, m = −l, −l + 1, . . . , l −1, l. First of all we observe
ψ
0,0
(x) =
1
π
3/2
e
−
x
2
2
∈ H
0,0
. (8.31)
Next, we note that (8.21) implies
[L
3
, x
±
] = ±x
±
, x
±
= x
1
±ix
2
,
[L
±
, x
±
] = 0, [L
±
, x
∓
] = ±2x
3
,
[L
2
, x
±
] = 2x
±
(1 ±L
3
) ∓2x
3
L
±
. (8.32)
Hence if ψ ∈ H
l,l
, then (x
1
±ix
2
)ψ ∈ H
l±1,l±1
. And thus
ψ
l,l
(x) =
1
√
l!
(x
1
±ix
2
)
l
ψ
0,0
(x) ∈ H
l,l
, (8.33)
respectively
ψ
l,m
(x) =
(l +m)!
(l −m)!(2l)!
L
l−m
−
ψ
l,l
(x) ∈ H
l,m
. (8.34)
The constants are chosen such that ψ
l,m
 = 1.
In summary,
Theorem 8.3. There exists an orthonormal basis of simultaneous eigenvec
tors for the operators L
2
and L
j
. Moreover, their spectra are given by
σ(L
2
) = ¦l(l + 1)[l ∈ N
0
¦, σ(L
3
) = Z. (8.35)
We will rederive this result using diﬀerent methods in Section 10.3.
154 8. Algebraic methods
8.3. The harmonic oscillator
Finally, let us consider another important model whose algebraic structure
is similar to those of the angular momentum, the harmonic oscillator
H = H
0
+ω
2
x
2
, ω > 0. (8.36)
As domain we will choose
D(H) = D = span¦x
α
e
−
x
2
2
[ α ∈ N
3
0
¦ ⊆ L
2
(R
3
) (8.37)
from our previous section.
We will ﬁrst consider the onedimensional case. Introducing
A
±
=
1
√
2
√
ωx ∓
1
√
ω
d
dx
(8.38)
we have
[A
−
, A
+
] = 1 (8.39)
and
H = ω(2N + 1), N = A
+
A
−
, (8.40)
for any function in D.
Moreover, since
[N, A
±
] = ±A
±
, (8.41)
we see that Nψ = nψ implies NA
±
ψ = (n ±1)A
±
ψ. Moreover, A
+
ψ
2
=
'ψ, A
−
A
+
ψ` = (n + 1)ψ
2
respectively A
−
ψ
2
= nψ
2
in this case and
hence we conclude that σ(N) ⊆ N
0
If Nψ
0
= 0, then we must have A
−
ψ = 0 and the normalized solution
of this last equation is given by
ψ
0
(x) =
ω
π
1/4
e
−
ωx
2
2
∈ D. (8.42)
Hence
ψ
n
(x) =
1
√
n!
A
n
+
ψ
0
(x) (8.43)
is a normalized eigenfunction of N corresponding to the eigenvalue n. More
over, since
ψ
n
(x) =
1
√
2
n
n!
ω
π
1/4
H
n
(
√
ωx)e
−
ωx
2
2
(8.44)
where H
n
(x) is a polynomial of degree n given by
H
n
(x) = e
x
2
2
x −
d
dx
n
e
−
x
2
2
= (−1)
n
e
x
2 d
n
dx
n
e
−x
2
, (8.45)
we conclude span¦ψ
n
¦ = D. The polynomials H
n
(x) are called Hermite
polynomials.
In summary,
8.3. The harmonic oscillator 155
Theorem 8.4. The harmonic oscillator H is essentially self adjoint on D
and has an orthonormal basis of eigenfunctions
ψ
n
1
,n
2
,n
3
(x) = ψ
n
1
(x
1
)ψ
n
2
(x
2
)ψ
n
3
(x
3
), (8.46)
with ψ
n
j
(x
j
) from (8.44). The spectrum is given by
σ(H) = ¦(2n + 3)ω[n ∈ N
0
¦. (8.47)
Finally, there is also a close connection with the Fourier transformation.
without restriction we choose ω = 1 and consider only one dimension. Then
it easy to verify that H commutes with the Fourier transformation
TH = HT. (8.48)
Moreover, by TA
±
= ∓iA
±
T we even infer
Tψ
n
=
1
√
n!
TA
n
+
ψ
0
=
(−i)
n
√
n!
A
n
+
Tψ
0
= (−i)
n
ψ
n
, (8.49)
since Tψ
0
= ψ
0
by Lemma 7.3. In particular,
σ(T) = ¦z ∈ C[z
4
= 1¦. (8.50)
Problem 8.2. Show that H = −
d
2
dx
2
+q can be written as H = AA
∗
, where
A = −
d
dx
+φ, if the diﬀerential equation ψ
+qψ = 0 has a positive solution.
Compute
˜
H = A
∗
A. (Hint: φ =
ψ
ψ
.)
Chapter 9
One dimensional
Schr¨odinger operators
9.1. SturmLiouville operators
In this section we want to illustrate some of the results obtained thus far by
investigating a speciﬁc example, the SturmLiouville equations.
τf(x) =
1
r(x)
−
d
dx
p(x)
d
dx
f(x) +q(x)f(x)
, f, pf
∈ AC
loc
(I) (9.1)
The case p = r = 1 can be viewed as the model of a particle in one
dimension in the external potential q. Moreover, the case of a particle in
three dimensions can in some situations be reduced to the investigation of
SturmLiouville equations. In particular, we will see how this works when
explicitly solving the hydrogen atom.
The suitable Hilbert space is
L
2
((a, b), r(x)dx), 'f, g` =
b
a
f(x)
∗
g(x)r(x)dx, (9.2)
where I = (a, b) ⊂ R is an arbitrary open interval.
We require
(i) p
−1
∈ L
1
loc
(I), realvalued
(ii) q ∈ L
1
loc
(I), realvalued
(iii) r ∈ L
1
loc
(I), positive
157
158 9. One dimensional Schr¨odinger operators
If a is ﬁnite and if p
−1
, q, r ∈ L
1
((a, c)) (c ∈ I), then the SturmLiouville
equation (9.1) is called regular at a. Similarly for b. If it is both regular at
a and b it is called regular.
The maximal domain of deﬁnition for τ in L
2
(I, r dx) is given by
D(τ) = ¦f ∈ L
2
(I, r dx)[f, pf
∈ AC
loc
(I), τf ∈ L
2
(I, r dx)¦. (9.3)
It is not clear that D(τ) is dense unless (e.g.) p ∈ AC
loc
(I), p
, q ∈ L
2
loc
(I),
r
−1
∈ L
∞
loc
(I) since C
∞
0
(I) ⊂ D(τ) in this case. We will defer the general
case to Lemma 9.4 below.
Since we are interested in selfadjoint operators H associated with (9.1),
we perform a little calculation. Using integration by parts (twice) we obtain
(a < c < d < b):
d
c
g
∗
(τf) rdy = W
d
(g
∗
, f) −W
c
(g
∗
, f) +
d
c
(τg)
∗
f rdy, (9.4)
for f, g, pf
, pg
∈ AC
loc
(I) where
W
x
(f
1
, f
2
) =
p(f
1
f
2
−f
1
f
2
)
(x) (9.5)
is called the modiﬁed Wronskian.
Equation (9.4) also shows that the Wronskian of two solutions of τu = zu
is constant
W
x
(u
1
, u
2
) = W(u
1
, u
2
), τu
1,2
= zu
1,2
. (9.6)
Moreover, it is nonzero if and only if u
1
and u
2
are linearly independent
(compare Theorem 9.1 below).
If we choose f, g ∈ D(τ) in (9.4), than we can take the limits c → a and
d → b, which results in
'g, τf` = W
b
(g
∗
, f) −W
a
(g
∗
, f) +'τg, f`, f, g ∈ D(τ). (9.7)
Here W
a,b
(g
∗
, f) has to be understood as limit.
Finally, we recall the following wellknown result from ordinary diﬀer
ential equations.
Theorem 9.1. Suppose rg ∈ L
1
loc
(I), then there exists a unique solution
f, pf
∈ AC
loc
(I) of the diﬀerential equation
(τ −z)f = g, z ∈ C, (9.8)
satisfying the initial condition
f(c) = α, (pf
)(c) = β, α, β ∈ C, c ∈ I. (9.9)
In addition, f is holomorphic with respect to z.
Note that f, pf
can be extended continuously to a regular end point.
9.1. SturmLiouville operators 159
Lemma 9.2. Suppose u
1
, u
2
are two solutions of (τ−z)u = 0 with W(u
1
, u
2
) =
1. Then any other solution of (9.8) can be written as (α, β ∈ C)
f(x) = u
1
(x)
α +
x
c
u
2
gr dy
+u
2
(x)
β −
x
c
u
1
gr dy
,
f
(x) = u
1
(x)
α +
x
c
u
2
gr dy
+u
2
(x)
β −
x
c
u
1
gr dy
. (9.10)
Note that the constants α, β coincide with those from Theorem 9.1 if u
1
(c) =
(pu
2
)(c) = 1 and (pu
1
)(c) = u
2
(c) = 0.
Proof. It suﬃces to check τf − z f = g. Diﬀerentiating the ﬁrst equation
of (9.10) gives the second. Next we compute
(pf
)
= (pu
1
)
α +
u
2
gr dy
+ (pu
2
)
β −
u
1
gr dy
−W(u
1
, u
2
)gr
= (q −z)u
1
α +
u
2
gr dy
+ (q −z)u
2
β −
u
1
g dy
−gr
= (q −z)f −gr (9.11)
which proves the claim.
Now we want to obtain a symmetric operator and hence we choose
A
0
f = τf, D(A
0
) = D(τ) ∩ AC
c
(I), (9.12)
where AC
c
(I) are the functions in AC(I) with compact support. This deﬁ
nition clearly ensures that the Wronskian of two such functions vanishes on
the boundary, implying that A
0
is symmetric. Our ﬁrst task is to compute
the closure of A
0
and its adjoint. For this the following elementary fact will
be needed.
Lemma 9.3. Suppose V is a vector space and l, l
1
, . . . , l
n
are linear func
tionals (deﬁned on all of V ) such that
¸
n
j=1
Ker(l
j
) ⊆ Ker(l). Then l =
¸
n
j=0
α
j
l
j
for some constants α
j
∈ C.
Proof. First of all it is no restriction to assume that the functionals l
j
are
linearly independent. Then the map L : V → C
n
, f → (l
1
(f), . . . , l
n
(f)) is
surjective (since x ∈ Ran(L)
⊥
implies
¸
n
j=1
x
j
l
j
(f) = 0 for all f). Hence
there are vectors f
k
∈ V such that l
j
(f
k
) = 0 for j = k and l
j
(f
j
) = 1. Then
f −
¸
n
j=1
l
j
(f)f
j
∈
¸
n
j=1
Ker(l
j
) and hence l(f)−
¸
n
j=1
l
j
(f)l(f
j
) = 0. Thus
we can choose α
j
= l(f
j
).
Now we are ready to prove
Lemma 9.4. The operator A
0
is densely deﬁned and its closure is given by
A
0
f = τf, D(A
0
) = ¦f ∈ D(τ) [ W
a
(f, g) = W
b
(f, g) = 0, ∀g ∈ D(τ)¦.
(9.13)
160 9. One dimensional Schr¨odinger operators
Its adjoint is given by
A
∗
0
f = τf, D(A
∗
0
) = D(τ). (9.14)
Proof. We start by computing A
∗
0
and ignore the fact that we don’t know
wether D(A
0
) is dense for now.
By (9.7) we have D(τ) ⊆ D(A
∗
0
) and it remains to show D(A
∗
0
) ⊆ D(τ).
If h ∈ D(A
∗
0
) we must have
'h, A
0
f` = 'k, f`, ∀f ∈ D(A
0
) (9.15)
for some k ∈ L
2
(I, r dx). Using (9.10) we can ﬁnd a
˜
h such that τ
˜
h = k and
from integration by parts we obtain
b
a
(h(x) −
˜
h(x))
∗
(τf)(x)r(x)dx = 0, ∀f ∈ D(A
0
). (9.16)
Clearly we expect that h −
˜
h will be a solution of the τu = 0 and to prove
this we will invoke Lemma 9.3. Therefore we consider the linear functionals
l(g) =
b
a
(h(x) −
˜
h(x))
∗
g(x)r(x)dx, l
j
(g) =
b
a
u
j
(x)
∗
g(x)r(x)dx, (9.17)
on L
2
c
(I, r dx), where u
j
are two solutions of τu = 0 with W(u
1
, u
2
) = 0.
We have Ker(l
1
) ∩ Ker(l
2
) ⊆ Ker(l). In fact, if g ∈ Ker(l
1
) ∩ Ker(l
2
), then
f(x) = u
1
(x)
x
a
u
2
(y)g(y)r(y)dy +u
2
(x)
b
x
u
1
(y)g(y)r(y)dy (9.18)
is in D(A
0
) and g = τf ∈ Ker(l) by (9.16). Now Lemma 9.3 implies
b
a
(h(x) −
˜
h(x) +α
1
u
1
(x) +α
2
u
2
(x))
∗
g(x)r(x)dx = 0, ∀g ∈ L
2
c
(I, rdx)
(9.19)
and hence h =
˜
h +α
1
u
1
+α
2
u
2
∈ D(τ).
Now what if D(A
0
) were not dense? Then there would be some freedom
in choice of k since we could always add a component in D(A
0
)
⊥
. So suppose
we have two choices k
1
= k
2
. Then by the above calculation, there are
corresponding functions
˜
h
1
and
˜
h
2
such that h =
˜
h
1
+ α
1,1
u
1
+ α
1,2
u
2
=
˜
h
2
+ α
2,1
u
1
+ α
2,2
u
2
. In particular,
˜
h
1
−
˜
h
2
is in the kernel of τ and hence
k
1
= τ
˜
h
1
= τ
˜
h
2
= k
2
contradiction our assumption.
Next we turn to A
0
. Denote the set on the right hand side of (9.13) by
D. Then we have D ⊆ D(A
∗∗
0
) = A
0
by (9.7). Conversely, since A
0
⊆ A
∗
0
we can use (9.7) to conclude
W
a
(f, h) +W
b
(f, h) = 0, f ∈ D(A
0
), h ∈ D(A
∗
0
). (9.20)
9.2. Weyl’s limit circle, limit point alternative 161
Now replace h by a
˜
h ∈ D(A
∗
0
) which coincides with h near a and vanishes
identically near b (Problem 9.1). Then W
a
(f, h) = W
a
(f,
˜
h) +W
b
(f,
˜
h) = 0.
Finally, W
b
(f, h) = −W
a
(f, h) = 0 shows f ∈ D.
Example. If τ is regular at a, then f ∈ D(A
0
) if and only if f(a) =
(pf
)(a) = 0. This follows since we can prescribe the values of g(a), (pg
)(a)
for g ∈ D(τ) arbitrarily.
This result shows that any selfadjoint extension of A
0
must lie between
A
0
and A
∗
0
. Moreover, selfadjointness seems to be related to the Wronskian
of two functions at the boundary. Hence we collect a few properties ﬁrst.
Lemma 9.5. Suppose v ∈ D(τ) with W
a
(v
∗
, v) = 0 and there is a
ˆ
f ∈ D(τ)
with W(v
∗
,
ˆ
f)
a
= 0. then we have
W
a
(v, f) = 0 ⇔ W
a
(v, f
∗
) = 0 ∀f ∈ D(τ) (9.21)
and
W
a
(v, f) = W
a
(v, g) = 0 ⇒ W
a
(g
∗
, f) = 0 ∀f, g ∈ D(τ) (9.22)
Proof. For all f
1
, . . . , f
4
∈ D(τ) we have the Pl¨ ucker identity
W
x
(f
1
, f
2
)W
x
(f
3
, f
4
) +W
x
(f
1
, f
3
)W
x
(f
4
, f
2
) +W
x
(f
1
, f
4
)W
x
(f
2
, f
3
) = 0
(9.23)
which remains valid in the limit x → a. Choosing f
1
= v, f
2
= f, f
3
=
v
∗
, f
4
=
ˆ
f we infer (9.21). Choosing f
1
= f, f
2
= g
∗
, f
3
= v, f
4
=
ˆ
f we
infer (9.22).
Problem 9.1. Given α, β, γ, δ, show that there is a function f ∈ D(τ)
restricted to [c, d] ⊆ (a, b) such that f(c) = α, (pf)(c) = β and f(d) = γ,
(pf)(c) = δ. (Hint: Lemma 9.2)
Problem 9.2. Let A
0
= −
d
2
dx
2
, D(A
0
) = ¦f ∈ H
2
[0, 1][f(0) = f(1) = 0¦.
and B = q, D(B) = ¦f ∈ L
2
(0, 1)[qf ∈ L
2
(0, 1)¦. Find a q ∈ L
1
(0, 1) such
that D(A
0
) ∩ D(B) = ¦0¦. (Hint: Problem 0.18)
9.2. Weyl’s limit circle, limit point alternative
We call τ limit circle (l.c.) at a if there is a v ∈ D(τ) with W
a
(v
∗
, v) = 0
such that W
a
(v, f) = 0 for at least one f ∈ D(τ). Otherwise τ is called
limit point (l.p.) at a. Similarly for b.
Example. If τ is regular at a, it is limit circle at a. Since
W
a
(v, f) = (pf
)(a)v(a) −(pv
)(a)f(a) (9.24)
any realvalued v with (v(a), (pv
)(a)) = (0, 0) works.
162 9. One dimensional Schr¨odinger operators
Note that if W
a
(f, v) = 0, then W
a
(f, Re(v)) = 0 or W
a
(f, Im(v)) = 0.
Hence it is no restriction to assume that v is real and W
a
(v
∗
, v) = 0 is
trivially satisﬁed in this case. In particular, τ is limit point if and only if
W
a
(f, g) = 0 for all f, g ∈ D(τ).
Theorem 9.6. If τ is l.c. at a, then let v ∈ D(τ) with W(v
∗
, v)
a
= 0 and
W(v, f)
a
= 0 for some f ∈ D(τ). Similarly, if τ is l.c. at b, let w be an
analogous function. Then the operator
A : D(A) → L
2
(I, r dx)
f → τf
(9.25)
with
D(A) = ¦f ∈ D(τ)[ W
a
(v, f) = 0 if l.c. at a
W
b
(w, f) = 0 if l.c. at b¦
(9.26)
is selfadjoint.
Proof. Clearly A ⊆ A
∗
⊆ A
∗
0
. Let g ∈ D(A
∗
). As in the computation of A
0
we conclude W
a
(f, g) = 0 for all f ∈ D(A). Moreover, we can choose f such
that it coincides with v near a and hence W
a
(v, g) = 0, that is g ∈ D(A).
The name limit circle respectively limit point stems from the original
approach of Weyl, who considered the set of solutions τu = zu, z ∈ C`R
which satisfy W
c
(u
∗
, u) = 0. They can be shown to lie on a circle which
converges to a circle respectively point as c → a (or c → b).
Example. If τ is regular at a we can choose v such that (v(a), (pv
)(a)) =
(sin(α), −cos(α)), α ∈ [0, π), such that
W
a
(v, f) = cos(α)f(a) + sin(α)(pf
)(a). (9.27)
The most common choice α = 0 is known as Dirichlet boundary condi
tion f(a) = 0.
Next we want to compute the resolvent of A.
Lemma 9.7. Suppose z ∈ ρ(A), then there exists a solution u
a
(z, x) which
is in L
2
((a, c), r dx) and which satisﬁes the boundary condition at a if τ
is l.c. at a. Similarly, there exists a solution u
b
(z, x) with the analogous
properties near b.
The resolvent of A is given by
(A−z)
−1
g(x) =
b
a
G(z, x, y)g(y)r(y)dy, (9.28)
where
G(z, x, y) =
1
W(u
b
(z), u
a
(z))
u
b
(z, x)u
a
(z, y) x ≥ y
u
a
(z, x)u
b
(z, y) x ≤ y
. (9.29)
9.2. Weyl’s limit circle, limit point alternative 163
Proof. Let g ∈ L
2
c
(I, r dx) be realvalued and consider f = (A − z)
−1
g ∈
D(A). Since (τ −z)f = 0 near a respectively b, we obtain u
a
(z, x) by setting
it equal to f near a and using the diﬀerential equation to extend it to the
rest of I. Similarly we obtain u
b
. The only problem is that u
a
or u
b
might
be identically zero. Hence we need to show that this can be avoided by
choosing g properly.
Fix z and let g be supported in (c, d) ⊂ I. Since (τ −z)f = g we have
f(x) = u
1
(x)
α +
x
a
u
2
gr dy
+u
2
(x)
β +
b
x
u
1
gr dy
. (9.30)
Near a (x < c) we have f(x) = αu
1
(x) +
˜
βu
2
(x) and near b (x > d) we have
f(x) = ˜ αu
1
(x) +βu
2
(x), where ˜ α = α +
b
a
u
2
gr dy and
˜
β = β +
b
a
u
1
gr dy.
If f vanishes identically near both a and b we must have α = β = ˜ α =
˜
β = 0
and thus α = β = 0 and
b
a
u
j
(y)g(y)r(y)dy = 0, j = 1, 2. This case can
be avoided choosing g suitable and hence there is at least one solution, say
u
b
(z).
Now choose u
1
= u
b
and consider the behavior near b. If u
2
is not square
integrable on (d, b), we must have β = 0 since βu
2
= f − ˜ αu
b
is. If u
2
is
square integrable, we can ﬁnd two functions in D(τ) which coincide with
u
b
and u
2
near b. Since W(u
b
, u
2
) = 1 we see that τ is l.c. at a and hence
0 = W
b
(u
b
, f) = W
b
(u
b
, ˜ αu
b
+ βu
2
) = β. Thus β = 0 in both cases and we
have
f(x) = u
b
(x)
α +
x
a
u
2
gr dy
+u
2
(x)
b
x
u
b
gr dy. (9.31)
Now choosing g such that
b
a
u
b
gr dy = 0 we infer existence of u
a
(z). Choos
ing u
2
= u
a
and arguing as before we see α = 0 and hence
f(x) = u
b
(x)
x
a
u
a
(y)g(y)r(y)dy +u
a
(x)
b
x
u
b
(y)g(y)r(y)dy
=
b
a
G(z, x, y)g(y)r(y)dy (9.32)
for any g ∈ L
2
c
(I, r dx). Since this set is dense the claim follows.
Example. If τ is regular at a with a boundary condition as in the previous
example, we can choose u
a
(z, x) to be the solution corresponding to the
initial conditions (u
a
(z, a), (pu
a
)(z, a)) = (sin(α), −cos(α)). In particular,
u
a
(z, x) exists for all z ∈ C.
If τ is regular at both a and b there is a corresponding solution u
b
(z, x),
again for all z. So the only values of z for which (A − z)
−1
does not exist
must be those with W(u
b
(z), u
a
(z)) = 0. However, in this case u
a
(z, x)
164 9. One dimensional Schr¨odinger operators
and u
b
(z, x) are linearly dependent and u
a
(z, x) = γu
b
(z, x) satisﬁes both
boundary conditions. That is, z is an eigenvalue in this case.
In particular, regular operators have pure point spectrum. We will see
in Theorem 9.10 below, that this holds for any operator which is l.c. at both
end points.
In the previous example u
a
(z, x) is holomorphic with respect to z and
satisﬁes u
a
(z, x)
∗
= u
a
(z
∗
, x) (since it coresponds to real initial conditions
and our diﬀerential equation has real coeﬃcients). In general we have:
Lemma 9.8. Suppose z ∈ ρ(A), then u
a
(z, x) from the previous lemma can
be chosen locally holomorphic with respect to z such that
u
a
(z, x)
∗
= u
a
(z
∗
, x). (9.33)
Similarly, for u
b
(z, x).
Proof. Since this is a local property near a we can suppose b is regular
and choose u
b
(z, x) such that (u
b
(z, b), (pu
b
)(z, b)) = (sin(β), −cos(β)) as in
the example above. In addition, choose a second solution v
b
(z, x) such that
(v
b
(z, b), (pv
b
)(z, b)) = (cos(β), sin(β)) and observe W(u
b
(z), v
b
(z)) = 1. If
z ∈ ρ(A), z is no eigenvalue and hence u
a
(z, x) cannot be a multiple of
u
b
(z, x). Thus we can set
u
a
(z, x) = v
b
(z, x) +m(z)u
b
(z, x) (9.34)
and it remains to show that m(z) is holomorphic with m(z)
∗
= m(z
∗
).
Choosing h with compact support in (a, c) and g with support in (c, b)
we have
'h, (A−z)
−1
g` = 'h, u
a
(z)`'g
∗
, u
b
(z)`
= ('h, v
b
(z)` +m(z)'h, u
b
(z)`)'g
∗
, u
b
(z)` (9.35)
(with a slight abuse of notation since u
b
, v
b
might not be square integrable).
Choosing (realvalued) functions h and g such that 'h, u
b
(z)`'g
∗
, u
b
(z)` = 0
we can solve for m(z):
m(z) =
'h, (A−z)
−1
g` −'h, v
b
(z)`'g
∗
, u
b
(z)`
'h, u
b
(z)`'g
∗
, u
b
(z)`
. (9.36)
This ﬁnishes the proof.
Example. We already know that τ = −
d
2
dx
2
on I = (−∞, ∞) gives rise to
the free Schr¨odinger operator H
0
. Furthermore,
u
±
(z, x) = e
∓
√
−zx
, z ∈ C, (9.37)
are two linearly independent solutions (for z = 0) and since Re(
√
−z) > 0
for z ∈ C`[0, ∞), there is precisely one solution (up to a constant multiple)
9.2. Weyl’s limit circle, limit point alternative 165
which is square integrable near ±∞, namely u
±
. In particular, the only
choice for u
a
is u
−
and for u
b
is u
+
and we get
G(z, x, y) =
1
2
√
−z
e
−
√
−zx−y
(9.38)
which we already found in Section 7.4.
If, as in the previous example, there is only one square integrable solu
tion, there is no choice for G(z, x, y). But since diﬀerent boundary condi
tions must give rise to diﬀerent resolvents, there is no room for boundary
conditions in this case. This indicates a connection between our l.c., l.p.
distinction and square integrability of solutions.
Theorem 9.9 (Weyl alternative). The operator τ is l.c. at a if and only if
for one z
0
∈ C all solutions of (τ − z
0
)u = 0 are square integrable near a.
This then holds for all z ∈ C. Similarly for b.
Proof. If all solutions are square integrable near a, τ is l.c. at a since the
Wronskian of two linearly independent solutions does not vanish.
Conversely, take two functions v, ˜ v ∈ D(τ) with W
a
(v, ˜ v) = 0. By con
sidering real and imaginary parts it is no restriction th assume that v and
˜ v are realvalued. Thus they give rise to two diﬀerent selfadjoint operators
A and
˜
A (choose any ﬁxed w for the other endpoint). Let u
a
and ˜ u
a
be the
corresponding solutions from above, then W(u
a
, ˜ u
a
) = 0 (since otherwise
A =
˜
A by Lemma 9.5) and thus there are two linearly independent solutions
which are square integrable near a. Since any other solution can be written
as a linear combination of those two, every solution is square integrable near
a.
It remains to show that all solutions of (τ − z)u = 0 for all z ∈ C are
square integrable near a if τ is l.c. at a. In fact, the above argument ensures
this for every z ∈ ρ(A) ∩ ρ(
˜
A), that is, at least for all z ∈ C`R.
Suppose (τ −z)u = 0. and choose two linearly independent solutions u
j
,
j = 1, 2, of (τ − z
0
)u = 0 with W(u
1
, u
2
) = 1. Using (τ − z
0
)u = (z − z
0
)u
and (9.10) we have (a < c < x < b)
u(x) = αu
1
(x) +βu
2
(x) + (z −z
0
)
x
c
(u
1
(x)u
2
(y) −u
1
(y)u
2
(x))u(y)r(y) dy.
(9.39)
Since u
j
∈ L
2
((c, b), rdx) we can ﬁnd a constant M ≥ 0 such that
b
c
[u
1,2
(y)[
2
r(y) dy ≤ M. (9.40)
166 9. One dimensional Schr¨odinger operators
Now choose c close to b such that [z − z
0
[M
2
≤ 1/4. Next, estimating the
integral using Cauchy–Schwarz gives
x
c
(u
1
(x)u
2
(y) −u
1
(y)u
2
(x))u(y)r(y) dy
2
≤
x
c
[u
1
(x)u
2
(y) −u
1
(y)u
2
(x)[
2
r(y) dy
x
c
[u(y)[
2
r(y) dy
≤ M
[u
1
(x)[
2
+[u
2
(x)[
2
x
c
[u(y)[
2
r(y) dy (9.41)
and hence
x
c
[u(y)[
2
r(y) dy ≤ ([α[
2
+[β[
2
)M + 2[z −z
0
[M
2
x
c
[u(y)[
2
r(y) dy
≤ ([α[
2
+[β[
2
)M +
1
2
x
c
[u(y)[
2
r(y) dy. (9.42)
Thus
x
c
[u(y)[
2
r(y) dy ≤ 2([α[
2
+[β[
2
)M (9.43)
and since u ∈ AC
loc
(I) we have u ∈ L
2
((c, b), r dx) for every c ∈ (a, b).
Note that all eigenvalues are simple. If τ is l.p. at one endpoint this is
clear, since there is at most one solution of (τ − λ)u = 0 which is square
integrable near this end point. If τ is l.c. this also follows since the fact that
two solutions of (τ − λ)u = 0 satisfy the same boundary condition implies
that their Wronskian vanishes.
Finally, led us shed some additional light on the number of possible
boundary conditions. Suppose τ is l.c. at a and let u
1
, u
2
be two solutions
of τu = 0 with W(u
1
, u
2
) = 1. Abbreviate
BC
j
x
(f) = W
x
(u
j
, f), f ∈ D(τ). (9.44)
Let v be as in Theorem 9.6, then, using Lemma 9.5 it is not hard to see that
W
a
(v, f) = 0 ⇔ cos(α)BC
1
a
(f) + sin(α)BC
2
a
(f) = 0, (9.45)
where tan(α) = −
BC
1
a
(v)
BC
2
a
(v)
. Hence all possible boundary conditions can be
parametrized by α ∈ [0, π). If τ is regular at a and if we choose u
1
(a) =
p(a)u
2
(a) = 1 and p(a)u
1
(a) = u
2
(a) = 0, then
BC
1
a
(f) = f(a), BC
2
a
(f) = p(a)f
(a) (9.46)
and the boundary condition takes the simple form
cos(α)f(a) + sin(α)p(a)f
(a) = 0. (9.47)
Finally, note that if τ is l.c. at both a and b, then Theorem 9.6 does not give
all possible selfadjoint extensions. For example, one could also choose
BC
1
a
(f) = e
iα
BC
1
b
(f), BC
2
a
(f) = e
iα
BC
2
b
(f). (9.48)
9.2. Weyl’s limit circle, limit point alternative 167
The case α = 0 gives rise to periodic boundary conditions in the regular
case.
Now we turn to the investigation of the spectrum of A. If τ is l.c. at
both endpoints, then the spectrum of A is very simple
Theorem 9.10. If τ is l.c. at both end points, then the resolvent is a Hilbert–
Schmidt operator, that is,
b
a
b
a
[G(z, x, y)[
2
r(y)dy r(x)dx < ∞. (9.49)
In particular, the spectrum of any self adjoint extensions is purely discrete
and the eigenfunctions (which are simple) form an orthonormal basis.
Proof. This follows from the estimate
b
a
x
a
[u
b
(x)u
a
(y)[
2
r(y)dy +
b
x
[u
b
(y)u
a
(x)[
2
r(y)dy
r(x)dx
≤ 2
b
a
[u
a
(y)[
2
r(y)dy
b
a
[u
b
(y)[
2
r(y)dy, (9.50)
which shows that the resolvent is Hilbert–Schmidt and hence compact.
If τ is not l.c. the situation is more complicated and we can only say
something about the essential spectrum.
Theorem 9.11. All self adjoint extensions have the same essential spec
trum. Moreover, if A
ac
and A
cb
are selfadjoint extensions of τ restricted to
(a, c) and (c, b) (for any c ∈ I), then
σ
ess
(A) = σ
ess
(A
ac
) ∪ σ
ess
(A
cb
). (9.51)
Proof. Since (τ −i)u = 0 has two linearly independent solutions, the defect
indices are at most two (they are zero if τ is l.p. at both end points, one if
τ is l.c. at one and l.p. at the other end point, and two if τ is l.c. at both
endpoints). Hence the ﬁrst claim follows from Theorem 6.19.
For the second claim restrict τ to the functions with compact support
in (a, c) ∪ (c, d). Then, this operator is the orthogonal sum of the operators
A
0,ac
and A
0,cb
. Hence the same is true for the adjoints and hence the defect
indices of A
0,ac
⊕ A
0,cb
are at most four. Now note that A and A
ac
⊕ A
cb
are both selfadjoint extensions of this operator. Thus the second claim also
follows from Theorem 6.19.
Problem 9.3. Compute the spectrum and the resolvent of τ = −
d
2
dx
2
, I =
(0, ∞) deﬁned on D(A) = ¦f ∈ D(τ)[f(0) = 0¦.
168 9. One dimensional Schr¨odinger operators
9.3. Spectral transformations
In this section we want to provide some fundamental tools for investigating
the spectra of SturmLiouville operators and, at the same time, give some
nice illustrations of the spectral theorem.
Example. Consider again τ = −
d
2
dx
2
on I = (−∞, ∞). From Section 7.2
we know that the Fourier transform maps the associated operator H
0
to the
multiplication operator with p
2
in L
2
(R). To get multiplication by λ, as in
the spectral theorem, we set p =
√
λ and split the Fourier integral into a
positive and negative part
(Uf)(λ) =
R
e
i
√
λx
f(x) dx
R
e
−i
√
λx
f(x) dx
, λ ∈ σ(H
0
) = [0, ∞). (9.52)
Then
U : L
2
(R) →
2
j=1
L
2
(R,
χ
[0,∞)
(λ)
2
√
λ
dλ) (9.53)
is the spectral transformation whose existence is guaranteed by the spectral
theorem (Lemma 3.3).
Note that in the previous example the kernel e
±i
√
λx
of the integral trans
form U is just a pair of linearly independent solutions of the underlying
diﬀerential equation (though no eigenfunctions, since they are not square
integrable).
More general, if
U : L
2
(I, r dx) → L
2
(R, dµ), f(x) →
R
u(λ, x)f(x)r(x) dx (9.54)
is an integral transformation which maps a selfadjoint SturmLiouville op
erator A to multiplication by λ, then its kernel u(λ, x) is a solution of the
underlying diﬀerential equation. This formally follows from UAf = λUf
which implies
R
u(λ, x)(τ −λ)f(x)r(x) dx =
R
(τ −λ)u(λ, x)f(x)r(x) dx (9.55)
and hence (τ −λ)u(λ, .) = 0.
Lemma 9.12. Suppose
U : L
2
(I, r dx) →
k
j=1
L
2
(R, dµ
j
) (9.56)
is a spectral mapping as in Lemma 3.3. Then U is of the form
Uf(x) =
b
a
u(λ, x)f(x)r(x) dx, (9.57)
9.3. Spectral transformations 169
where u(λ, x) = (u
1
(λ, x), . . . , u
k
(λ, x)) and each u
j
(λ, .) is a solution of
τu
j
= λu
j
for a.e. λ (with respect to µ
j
). The inverse is given by
U
−1
F(λ) =
k
¸
j=1
R
u
j
(λ, x)
∗
F
j
(λ)dµ
j
(λ). (9.58)
Moreover, the solutions u
j
(λ) are linearly independent if the spectral
measures are ordered and, if τ is l.c. at some endpoint, they satisfy the
boundary condition. In particular, for ordered spectral measures we have
always k ≤ 2 and even k = 1 if τ is l.c. at one endpoint.
Proof. Using U
j
R
A
(z) =
1
λ−z
U
j
we have
U
j
f(x) = (λ −z)U
j
b
a
G(z, x, y)f(y)r(y) dy. (9.59)
If we restrict R
A
(z) to a compact interval [c, d] ⊂ (a, b), then R
A
(z)χ
[c,d]
is Hilbert–Schmidt since G(z, x, y)χ
[c,d]
(y) is square integrable over (a, b)
(a, b). Hence U
j
χ
[c,d]
= (λ −z)U
j
R
A
(z)χ
[c,d]
is Hilbert–Schmidt as well and
by Lemma 6.9 there is a corresponding kernel u
[c,d]
j
(λ, y) such that
(U
j
χ
[c,d]
f)(λ) =
b
a
u
[c,d]
j
(λ, x)f(x)r(x) dx. (9.60)
Now take a larger compact interval [ˆ c,
ˆ
d] ⊇ [c, d], then the kernels coincide
on [c, d], u
[c,d]
j
(λ, .) = u
[ˆ c,
ˆ
d]
j
(λ, .)χ
[c,d]
, since we have U
j
χ
[c,d]
= U
j
χ
[ˆ c,
ˆ
d]
χ
[c,d]
.
In particular, there is a kernel u
j
(λ, x) such that
U
j
f(x) =
b
a
u
j
(λ, x)f(x)r(x) dx (9.61)
for every f with compact support in (a, b). Since functions with compact
support are dense and U
j
is continuous, this formula holds for any f (pro
vided the integral is understood as the corresponding limit).
Using the fact that U is unitary, 'F, Ug` = 'U
−1
F, g`, we see
¸
j
R
F
j
(λ)
∗
b
a
u
j
(λ, x)g(x)r(x) dx =
b
a
(U
−1
F)(x)
∗
g(x)r(x) dx. (9.62)
Interchanging integrals on the right hand side (which is permitted at least
for g, F with compact support), the formula for the inverse follows.
Next, from U
j
Af = λU
j
f we have
b
a
u
j
(λ, x)(τf)(x)r(x) dx = λ
b
a
u
j
(λ, x)f(x)r(x) dx (9.63)
170 9. One dimensional Schr¨odinger operators
for a.e. λ and every f ∈ D(A
0
). Restricting everything to [c, d] ⊂ (a, b)
the above equation implies u
j
(λ, .)[
[c,d]
∈ D(A
∗
cd,0
) and A
∗
ab,0
u
j
(λ, .)[
[c,d]
=
λu
j
(λ, .)[
[c,d]
. In particular, u
j
(λ, .) is a solution of τu
j
= λu
j
. Moreover, if
u
j
(λ, .) is τ is l.c. near a, we can choose α = a and f to satisfy the boundary
condition.
Finally, ﬁx l ≤ k. If we assume the µ
j
are ordered, there is a set Ω
l
such
that µ
j
(Ω
l
) = 0 for 1 ≤ j ≤ l. Suppose
l
¸
j=1
c
j
(λ)u
j
(λ, x) = 0 (9.64)
then we have
l
¸
j=1
c
j
(λ)F
j
(λ) = 0, F
j
= U
j
f, (9.65)
for every f. Since U is surjective, we can prescribe F
j
arbitrarily, e.g.,
F
j
(λ) = 1 for j = j
0
and F
j
(λ) = 0 else which shows c
j
0
(λ) = 0. Hence
u
j
(λ, x), 1 ≤ j ≤ l, are linearly independent for λ ∈ Ω
l
which shows k ≤
2 since there are at most two linearly independent solutions. If τ is l.c.
and u
j
(λ, x) must satisfy the boundary condition, there is only one linearly
independent solution and thus k = 1.
Please note that the integral in (9.57) has to be understood as
U
j
f(x) = lim
α↓a,β↑b
β
α
u
j
(λ, x)f(x)r(x) dx, (9.66)
where the limit is taken in L
2
(R, dµ
j
). Similarly for (9.58).
For simplicity we will only pursue the case where one endpoint, say a,
is regular. The general case can usually be reduced to this case by choosing
c ∈ (a, b) and splitting A as in Theorem 9.11.
We choose a boundary condition
cos(α)f(a) + sin(α)p(a)f
(a) = 0 (9.67)
and introduce two solution s(z, x) and c(z, x) of τu = zu satisfying the
initial conditions
s(z, a) = −sin(α), p(a)s
(z, a) = cos(α),
c(z, a) = cos(α), p(a)c
(z, a) = sin(α). (9.68)
Note that s(z, x) is the solution which satisﬁes the boundary condition at
a, that is, we can choose u
a
(z, x) = s(z, x). Moreover, in our previous
lemma we have u
1
(λ, x) = γ
a
(λ)s(λ, x) and using the rescaling dµ(λ) =
9.3. Spectral transformations 171
[γ
a
(λ)[
2
dµ
a
(λ) and (U
1
f)(λ) = γ
a
(λ)(Uf)(λ) we obtain a unitary map
U : L
2
(I, r dx) → L
2
(R, dµ), (Uf)(λ) =
b
a
s(λ, x)f(x)r(x)dx (9.69)
with inverse
(U
−1
F)(x) =
R
s(λ, x)F(λ)dµ(λ). (9.70)
Note however, that while this rescaling gets rid of the unknown factor γ
a
(λ),
it destroys the normalization of the measure µ. For µ
1
we know µ
1
(R) = 1,
but µ might not even be bounded! In fact, it turns out that µ is indeed
unbounded.
So up to this point we have our spectral transformation U which maps A
to multiplication by λ, but we know nothing about the measure µ. Further
more, the measure µ is the object of desire since it contains all the spectral
information of A. So our next aim must be to compute µ. If A has only
pure point spectrum (i.e., only eigenvalues) this is straightforward as the
following example shows.
Example. Suppose E ∈ σ
p
(A) is an eigenvalue. Then s(E, x) is the corre
sponding eigenfunction and hence the same is true for S
E
(λ) = (Us(E))(λ).
In particular, S
E
(λ) = 0 for a.e. λ = E, that is
S
E
(λ) =
s(E)
2
, λ = E
0, λ = 0
. (9.71)
Moreover, since U is unitary we have
s(E)
2
=
b
a
s(E, x)
2
r(x)dx =
R
S
E
(λ)
2
dµ(λ) = s(E)
4
µ(¦E¦), (9.72)
that is µ(¦E¦) = s(E)
−2
. In particular, if A has pure point spectrum
(e.g., if τ is limit circle at both endpoints), we have
dµ(λ) =
∞
¸
j=1
1
s(E
j
)
2
dΘ(λ −E
j
), σ
p
(A) = ¦E
j
¦
∞
j=1
, (9.73)
where dΘ is the Dirac measure centered at 0. For arbitrary A, the above
formula holds at least for the pure point part µ
pp
.
In the general case we have to work a bit harder. Since c(z, x) and s(z, x)
are linearly independent solutions,
W(c(z), s(z)) = 1, (9.74)
we can write u
b
(z, x) = γ
b
(z)(c(z, x) +m
b
(z)s(z, x)), where
m
b
(z) =
cos(α)p(a)u
b
(z, a) −sin(α)u
b
(z, a)
cos(α)u
b
(z, a) + sin(α)p(a)u
b
(z, a)
, z ∈ ρ(A), (9.75)
172 9. One dimensional Schr¨odinger operators
is known as WeylTitchmarsh mfunction. Note that m
b
(z) is holomor
phic in ρ(A) and that
m
b
(z)
∗
= m
b
(z
∗
) (9.76)
since the same is true for u
b
(z, x) (the denominator in (9.75) only vanishes if
u
b
(z, x) satisﬁes the boundary condition at a, that is, if z is an eigenvalue).
Moreover, the constant γ
b
(z) is of no importance and can be chosen equal
to one,
u
b
(z, x) = c(z, x) +m
b
(z)s(z, x). (9.77)
Lemma 9.13. The Weyl mfunction is a Herglotz function and satisﬁes
Im(m
b
(z)) = Im(z)
b
a
[u
b
(z, x)[
2
r(x) dx, (9.78)
where u
b
(z, x) is normalized as in (9.77).
Proof. Given two solutions u(x), v(x) of τu = zu, τv = ˆ zv it is straightfor
ward to check
(ˆ z −z)
x
a
u(y)v(y)r(y) dy = W
x
(u, v) −W
a
(u, v) (9.79)
(clearly it is true for x = a, now diﬀerentiate with respect to x). Now choose
u(x) = u
b
(z, x) and v(x) = u
b
(z, x)
∗
= u
b
(z
∗
, x),
−2Im(z)
x
a
[u
b
(z, y)[
2
r(y) dy = W
x
(u
b
(z), u
b
(z)
∗
) −2Im(m
b
(z)), (9.80)
and observe that W
x
(u
b
, u
∗
b
) vanishes as x ↑ b, since both u
b
and u
∗
b
are in
D(τ) near b.
Lemma 9.14. We have
(Uu
b
(z))(λ) =
1
λ −z
, (9.81)
where u
b
(z, x) is normalized as in (9.77).
Proof. First of all note that from R
A
(z)f = U
−1 1
λ−z
Uf we have
b
a
G(z, x, y)f(y)r(y) dy =
R
s(λ, x)F(λ)
λ −z
dµ(λ), (9.82)
where F = Uf. Here equality is to be understood in L
2
, that is for a.e. x.
However, the right hand side is continuous with respect to x and so is the
left hand side, at least if F has compact support. Hence in this case the
formula holds for all x and we can choose x = a to obtain
sin(α)
b
a
u
b
(z, y)f(y)r(y) dy = sin(α)
R
F(λ)
λ −z
dµ(λ), (9.83)
9.3. Spectral transformations 173
for all f, where F has compact support. Since these functions are dense, the
claim follows if we can cancel sin(α), that is, α = 0. To see the case α = 0,
ﬁrst diﬀerentiate (9.82) with respect to x before setting x = a.
Now combining the last two lemmas we infer from unitarity of U that
Im(m
b
(z)) = Im(z)
b
a
[u
b
(z, x)[
2
r(x) dx = Im(z)
R
1
[λ −z[
2
dµ(λ) (9.84)
and since a holomorphic function is determined up to a real constant by its
imaginary part we obtain
Theorem 9.15. The Weyl mfunction is given by
m
b
(z) = d +
R
1
λ −z
−
λ
1 +λ
2
dµ(λ), d ∈ R, (9.85)
and
d = Re(m
b
(i)),
R
1
1 +λ
2
dµ(λ) = Im(m
b
(i)) < ∞. (9.86)
Moreover, µ is given by Stieltjes inversion formula
µ(λ) = lim
δ↓0
lim
ε↓0
1
π
λ+δ
δ
Im(m
b
(λ + iε))dλ, (9.87)
where
Im(m
b
(λ + iε)) = ε
b
a
[u
b
(λ + iε, x)[
2
r(x) dx. (9.88)
Proof. Choosing z = i in (9.84) shows (9.86) and hence the right hand side
of (9.85) is a welldeﬁned holomorphic function in C`R. By Im(
1
λ−z
−
λ
1+λ
2
) =
Im(z)
λ−z
2
its imaginary part coincides with that of m
b
(z) and hence equality
follows. Stieltjes inversion formula follows as in the case where the measure
is bounded.
Example. Consider τ = −
d
2
dx
2
on I = (0, ∞). Then
c(z, x) = cos(α) cos(
√
zx) +
sin(α)
√
z
sin(
√
zx) (9.89)
and
s(z, x) = −sin(α) cos(
√
zx) +
cos(α)
√
z
sin(
√
zx). (9.90)
Moreover,
u
b
(z, x) = u
b
(z, 0)e
−
√
−zx
(9.91)
and thus
m
b
(z) =
√
−z cos(α) + sin(α)
√
−z sin(α) −cos(α)
(9.92)
174 9. One dimensional Schr¨odinger operators
respectively
dµ(λ) =
√
λ
π(cos(α)
2
+λsin(α)
2
)
dλ. (9.93)
Note that if α = 0 we even have
1
λ−z
dµ(λ) < 0 in the previous example
and hence
m
b
(z) = cot(α) +
R
1
λ −z
dµ(λ) (9.94)
in this case (the factor −cot(α) follows by considering the limit [z[ → ∞
of both sides). One can show that this remains true in the general case.
Formally it follows by choosing x = a in u
b
(z, x) = (U
−1 1
λ−z
)(x), however,
since we know equality only for a.e. x, a more careful analysis is needed.
We conclude this section with some asymptotics for large z away from
the spectrum. For simplicity we only consider the case p = k ≡ 1. We ﬁrst
observe (Problem 9.4):
Lemma 9.16. We have
c(z, x) =
1
2
cos(α) +
sin(α)
√
−z
e
√
−z(x−a)
(1 +O(
1
√
−z
))
s(z, x) =
1
2
−sin(α) +
cos(α)
√
−z
e
√
−z(x−a)
(1 +O(
1
√
−z
)) (9.95)
for ﬁxed x ∈ (a, b) as Im(z) → ∞.
Furthermore we have
Lemma 9.17. The Green’s function of A
G(z, x, y) = s(z, x)u
b
(z, y), x ≤ y, (9.96)
satisﬁes
lim
Im(z)→∞
G(z, x, y) = 0 (9.97)
for every x, y ∈ (a, b).
Proof. Let ϕ be some continuous positive function supported in [−1, 1] with
ϕ(x)dx = 1 and set ϕ
k
(x) = k ϕ(k x). Then, by R
A
(z) ≤ Im(z)
−1
(see
Theorem 2.14),
[G(z, x, y)[ ≤ [G
k
(z, x, y)[ +[G(z, x, y) −G
k
(z, x, y)[
≤
kϕ
2
Im(z)
+[G(z, x, y) −G
k
(z, x, y)[, (9.98)
where
G
k
(z, x, y) = 'ϕ
k
(. −x), R
A
(z)ϕ
k
(. −y)`. (9.99)
9.3. Spectral transformations 175
Since G(z, x, y) = lim
k→∞
G
k
(z, x, y) we can ﬁrst choose k suﬃciently large
such that [G(z, x, y) − G
k
(z, x, y)[ ≤
ε
2
and then Im(z) so large such that
kϕ
2
Im(z)
≤
ε
2
. Hence [G(z, x, y)[ ≤ ε an the claim follows.
Combining the last two lemmas we also obtain
Lemma 9.18. The Weyl mfunction satisﬁes
m
b
(z) =
cot(α) +O(
1
√
−z
), α = 0
−
√
−z +O(
1
√
−z
), α = 0
(9.100)
as Im(z) → ∞.
Proof. Plugging the asymptotic expansions from Lemma 9.16 into Lemma 9.17
we see
m
b
(z) = −
c(z, x)
s(z, x)
+o(e
−2
√
−z(x−a)
) (9.101)
form which the claim follows.
Note that assuming q ∈ C
k
([a, b)) one can obtain further asymptotic
terms in Lemma 9.16 and hence also in the expansion of m
b
(z).
Problem 9.4. Prove the asymptotics for c(z, x) and s(z, x) from Lemma 9.16.
(Hint: Use that
c(z, x) = cos(α) cosh(
√
−zx) +
sin(α)
√
−z
sinh(
√
−zx)
−
1
√
−z
x
a
sinh(
√
−z(x −y))q(y)c(z, y)dy
by Lemma 9.2 and consider ˜ c(z, x) = e
−
√
−zx
c(z, x).)
Chapter 10
Oneparticle
Schr¨odinger operators
10.1. Selfadjointness and spectrum
Our next goal is to apply these results to Schr¨odinger operators. The Hamil
tonian of one particle in d dimensions is given by
H = H
0
+V, (10.1)
where V : R
d
→ R is the potential energy of the particle. We are mainly
interested in the case 1 ≤ d ≤ 3 and want to ﬁnd classes of potentials which
are relatively bounded respectively relatively compact. To do this we need
a better understanding of the functions in the domain of H
0
.
Lemma 10.1. Suppose n ≤ 3 and ψ ∈ H
2
(R
n
). Then ψ ∈ C
∞
(R
n
) and for
any a > 0 there is a b > 0 such that
ψ
∞
≤ aH
0
ψ +bψ. (10.2)
Proof. The important observation is that (p
2
+ γ
2
)
−1
∈ L
2
(R
n
) if n ≤ 3.
Hence, since (p
2
+γ
2
)
ˆ
ψ ∈ L
2
(R
n
), the CauchySchwarz inequality

ˆ
ψ
1
= (p
2
+γ
2
)
−1
(p
2
+γ
2
)
ˆ
ψ(p)
1
≤ (p
2
+γ
2
)
−1
 (p
2
+γ
2
)
ˆ
ψ(p). (10.3)
shows
ˆ
ψ ∈ L
1
(R
n
). But now everything follows from the RiemannLebesgue
lemma
ψ
∞
≤ (2π)
−n/2
(p
2
+γ
2
)
−1
(p
2
ˆ
ψ(p) +γ
2

ˆ
ψ(p))
= (γ/2π)
n/2
(p
2
+ 1)
−1
(γ
−2
H
0
ψ +ψ) (10.4)
177
178 10. Oneparticle Schr¨odinger operators
ﬁnishes the proof.
Now we come to our ﬁrst result.
Theorem 10.2. Let V be realvalued and V ∈ L
∞
∞
(R
n
) if n > 3 and V ∈
L
∞
∞
(R
n
) +L
2
(R
n
) if n ≤ 3. Then V is relatively compact with respect to H
0
.
In particular,
H = H
0
+V, D(H) = H
2
(R
n
), (10.5)
is selfadjoint, bounded from below and
σ
ess
(H) = [0, ∞). (10.6)
Moreover, C
∞
0
(R
n
) is a core for H.
Proof. Our previous lemma shows D(H
0
) ⊆ D(V ) and the rest follows
from Lemma 7.10 using f(p) = (p
2
− z)
−1
and g(x) = V (x). Note that
f ∈ L
∞
∞
(R
n
) ∩ L
2
(R
n
) for n ≤ 3.
Observe that since C
∞
c
(R
n
) ⊆ D(H
0
), we must have V ∈ L
2
loc
(R
n
) if
D(V ) ⊆ D(H
0
).
10.2. The hydrogen atom
We begin with the simple model of a single electron in R
3
moving in the
external potential V generated by a nucleus (which is assumed to be ﬁxed
at the origin). If one takes only the electrostatic force into account, then
V is given by the Coulomb potential and the corresponding Hamiltonian is
given by
H
(1)
= −∆−
γ
[x[
, D(H
(1)
) = H
2
(R
3
). (10.7)
If the potential is attracting, that is, if γ > 0, then it describes the hydrogen
atom and is probably the most famous model in quantum mechanics.
As domain we have chosen D(H
(1)
) = D(H
0
) ∩ D(
1
x
) = D(H
0
) and by
Theorem 10.2 we conclude that H
(1)
is selfadjoint. Moreover, Theorem 10.2
also tells us
σ
ess
(H
(1)
) = [0, ∞) (10.8)
and that H
(1)
is bounded from below
E
0
= inf σ(H
(1)
) > −∞. (10.9)
If γ ≤ 0 we have H
(1)
≥ 0 and hence E
0
= 0, but if γ > 0, we might have
E
0
< 0 and there might be some discrete eigenvalues below the essential
spectrum.
10.2. The hydrogen atom 179
In order to say more about the eigenvalues of H
(1)
we will use the fact
that both H
0
and V
(1)
= −γ/[x[ have a simple behavior with respect to
scaling. Consider the dilation group
U(s)ψ(x) = e
−ns/2
ψ(e
−s
x), s ∈ R, (10.10)
which is a strongly continuous oneparameter unitary group. The generator
can be easily computed
Dψ(x) =
1
2
(xp +px)ψ(x) = (xp −
in
2
)ψ(x), ψ ∈ o(R
n
). (10.11)
Now let us investigate the action of U(s) on H
(1)
H
(1)
(s) = U(−s)H
(1)
U(s) = e
−2s
H
0
+ e
−s
V
(1)
, D(H
(1)
(s)) = D(H
(1)
).
(10.12)
Now suppose Hψ = λψ, then
'ψ, [U(s), H]ψ` = 'U(−s)ψ, Hψ` −'Hψ, U(s)ψ` = 0 (10.13)
and hence
0 = lim
s→0
1
s
'ψ, [U(s), H]ψ` = lim
s→0
'U(−s)ψ,
H −H(s)
s
ψ`
= 'ψ, (2H
0
+V
(1)
)ψ`. (10.14)
Thus we have proven the virial theorem.
Theorem 10.3. Suppose H = H
0
+ V with U(−s)V U(s) = e
−s
V . Then
any normalized eigenfunction ψ corresponding to an eigenvalue λ satisﬁes
λ = −'ψ, H
0
ψ` =
1
2
'ψ, V ψ`. (10.15)
In particular, all eigenvalues must be negative.
This result even has some further consequences for the point spectrum
of H
(1)
.
Corollary 10.4. Suppose γ > 0. Then
σ
p
(H
(1)
) = σ
d
(H
(1)
) = ¦E
j−1
¦
j∈N
0
, E
0
< E
j
< E
j+1
< 0, (10.16)
with lim
j→∞
E
j
= 0.
Proof. Choose ψ ∈ C
∞
c
(R`¦0¦) and set ψ(s) = U(−s)ψ. Then
'ψ(s), H
(1)
ψ(s)` = e
−2s
'ψ, H
0
ψ` + e
−s
'ψ, V
(1)
ψ` (10.17)
which is negative for s large. Now choose a sequence s
n
→ ∞ such that
we have supp(ψ(s
n
)) ∩ supp(ψ(s
m
)) = ∅ for n = m. Then Theorem 4.11
(i) shows that rank(P
H
(1) ((−∞, 0))) = ∞. Since each eigenvalue E
j
has
ﬁnite multiplicity (it lies in the discrete spectrum) there must be an inﬁnite
number of eigenvalues which accumulate at 0.
180 10. Oneparticle Schr¨odinger operators
If γ ≤ 0 we have σ
d
(H
(1)
) = ∅ since H
(1)
≥ 0 in this case.
Hence we have gotten a quite complete picture of the spectrum of H
(1)
.
Next, we could try to compute the eigenvalues of H
(1)
(in the case γ > 0) by
solving the corresponding eigenvalue equation, which is given by the partial
diﬀerential equation
−∆ψ(x) −
γ
[x[
ψ(x) = λψ(x). (10.18)
For a general potential this is hopeless, but in our case we can use the rota
tional symmetry of our operator to reduce our partial diﬀerential equation
to ordinary ones.
First of all, it suggests itself to switch to spherical coordinates (x
1
, x
2
, x
3
) →
(r, θ, ϕ)
x
1
= r sin(θ) cos(ϕ), x
2
= r sin(θ) sin(ϕ), x
3
= r cos(θ), (10.19)
which correspond to a unitary transform
L
2
(R
3
) → L
2
((0, ∞), r
2
dr) ⊗L
2
((0, π), sin(θ)dθ) ⊗L
2
((0, 2π), dϕ). (10.20)
In these new coordinates (r, θ, ϕ) our operator reads
H
(1)
= −
1
r
2
∂
∂r
r
2
∂
∂r
+
1
r
2
L
2
+V (r), V (r) = −
γ
r
, (10.21)
where
L
2
= L
2
1
+L
2
2
+L
2
3
= −
1
sin(θ)
∂
∂θ
sin(θ)
∂
∂θ
−
1
sin(θ)
2
∂
2
∂ϕ
2
. (10.22)
(Recall the angular momentum operators L
j
from Section 8.2.)
Making the product ansatz (separation of variables)
ψ(r, θ, ϕ) = R(r)Θ(θ)Φ(ϕ) (10.23)
we obtain the following three SturmLiouville equations
−
1
r
2
d
dr
r
2
d
dr
+
l(l + 1)
r
2
+V (r)
R(r) = λR(r)
1
sin(θ)
−
d
dθ
sin(θ)
d
dθ
+
m
2
sin(θ)
Θ(θ) = l(l + 1)Θ(θ)
−
d
2
dϕ
2
Φ(ϕ) = m
2
Φ(ϕ) (10.24)
The form chosen for the constants l(l + 1) and m
2
is for convenience later
on. These equations will be investigated in the following sections.
10.3. Angular momentum 181
10.3. Angular momentum
We start by investigating the equation for Φ(ϕ) which associated with the
StumLiouville equation
τΦ = −Φ
, I = (0, 2π). (10.25)
since we want ψ deﬁned via (10.23) to be in the domain of H
0
(in particu
lar continuous), we choose periodic boundary conditions the StumLiouville
equation
AΦ = τΦ, D(A) = ¦Φ ∈ L
2
(0, 2π)[ Φ ∈ AC
1
[0, 2π],
Φ(0) = Φ(2π), Φ
(0) = Φ
(2π)¦
.
(10.26)
From our analysis in Section 9.1 we immediately obtain
Theorem 10.5. The operator A deﬁned via (10.25) is selfadjoint. Its
spectrum is purely discrete
σ(A) = σ
d
(A) = ¦m
2
[m ∈ Z¦ (10.27)
and the corresponding eigenfunctions
Φ
m
(ϕ) =
1
√
2π
e
imϕ
, m ∈ Z, (10.28)
form an orthonormal basis for L
2
(0, 2π).
Note that except for the lowest eigenvalue, all eigenvalues are twice de
generate.
We note that this operator is essentially the square of the angular mo
mentum in the third coordinate direction, since in polar coordinates
L
3
=
1
i
∂
∂ϕ
. (10.29)
Now we turn to the equation for Θ(θ)
τ
m
Θ(θ) =
1
sin(θ)
−
d
dθ
sin(θ)
d
dθ
+
m
2
sin(θ)
Θ(θ), I = (0, π), m ∈ N
0
.
(10.30)
For the investigation of the corresponding operator we use the unitary
transform
L
2
((0, π), sin(θ)dθ) → L
2
((−1, 1), dx), Θ(θ) → f(x) = Θ(arccos(x)).
(10.31)
The operator τ transforms to the somewhat simpler form
τ
m
= −
d
dx
(1 −x
2
)
d
dx
−
m
2
1 −x
2
. (10.32)
182 10. Oneparticle Schr¨odinger operators
The corresponding eigenvalue equation
τ
m
u = l(l + 1)u (10.33)
is the associated Legendre equation. For l ∈ N
0
it is solved by the
associated Legendre functions
P
lm
(x) = (1 −x
2
)
m/2
d
m
dx
m
P
l
(x), (10.34)
where
P
l
(x) =
1
2
l
l!
d
l
dx
l
(x
2
−1)
l
(10.35)
are the Legendre polynomials. This is straightforward to check. More
over, note that P
l
(x) are (nonzero) polynomials of degree l. A second,
linearly independent solution is given by
Q
lm
(x) = P
lm
(x)
x
0
dt
(1 −t
2
)P
lm
(t)
2
. (10.36)
In fact, for every SturmLiouville equation v(x) = u(x)
x
dt
p(t)u(t)
2
satisﬁes
τv = 0 whenever τu = 0. Now ﬁx l = 0 and note P
0
(x) = 1. For m = 0 we
have Q
00
= arctanh(x) ∈ L
2
and so τ
0
is l.c. at both end points. For m > 0
we have Q
0m
= (x±1)
−m/2
(C +O(x±1)) which shows that it is not square
integrable. Thus τ
m
is l.c. for m = 0 and l.p. for m > 0 at both endpoints.
In order to make sure that the eigenfunctions for m = 0 are continuous (such
that ψ deﬁned via (10.23) is continuous) we choose the boundary condition
generated by P
0
(x) = 1 in this case
A
m
f = τf, D(A
m
) = ¦f ∈ L
2
(−1, 1)[ f ∈ AC
1
(0, π), τf ∈ L
2
(−1, 1)
lim
x→±1
(1 −x
2
)f
(x) = 0¦
.
(10.37)
Theorem 10.6. The operator A
m
, m ∈ N
0
, deﬁned via (10.37) is self
adjoint. Its spectrum is purely discrete
σ(A
m
) = σ
d
(A
m
) = ¦l(l + 1)[l ∈ N
0
, l ≥ m¦ (10.38)
and the corresponding eigenfunctions
u
lm
(x) =
2l + 1
2
(l +m)!
(l −m)!
P
lm
(x), l ∈ N
0
, l ≥ m, (10.39)
form an orthonormal basis for L
2
(−1, 1).
Proof. By Theorem 9.6, A
m
is selfadjoint. Moreover, P
lm
is an eigenfunc
tion corresponding to the eigenvalue l(l +1) and it suﬃces to show that P
lm
form a basis. To prove this, it suﬃces to show that the functions P
lm
(x)
are dense. Since (1 − x
2
) > 0 for x ∈ (−1, 1) it suﬃces to show that the
functions (1 − x
2
)
−m/2
P
lm
(x) are dense. But the span of these functions
10.3. Angular momentum 183
contains every polynomial. Every continuous function can be approximated
by polynomials (in the sup norm and hence in the L
2
norm) and since the
continuous functions are dense, so are the polynomials.
The only thing remaining is the normalization of the eigenfunctions,
which can be found in any book on special functions.
Returning to our original setting we conclude that
Θ
lm
(θ) =
2l + 1
2
(l +m)!
(l −m)!
P
lm
(cos(θ)), l = m, m+ 1, . . . (10.40)
form an orthonormal basis for L
2
((0, π), sin(θ)dθ) for any ﬁxed m ∈ N
0
.
Theorem 10.7. The operator L
2
on L
2
((0, π), sin(θ)dθ) ⊗ L
2
((0, 2π)) has
a purely discrete spectrum given
σ(L
2
) = ¦l(l + 1)[l ∈ N
0
¦. (10.41)
The spherical harmonics
Y
lm
(θ, ϕ) = Θ
lm
(θ)Φ
m
(ϕ) =
2l + 1
4π
(l +[m[)!
(l −[m[)!
P
lm
(cos(θ))e
imϕ
, [m[ ≤ l,
(10.42)
form an orthonormal basis and satisfy L
2
Y
lm
= l(l + 1)Y
lm
and L
3
Y
lm
=
mY
lm
.
Proof. Everything follows from our construction, if we can show that Y
lm
form a basis. But this follows as in the proof of Lemma 1.9.
Note that transforming Y
lm
back to cartesian coordinates gives
Y
l,±m
(x) =
2l + 1
4π
(l +m)!
(l −m)!
˜
P
lm
(
x
3
r
)
x
1
±ix
2
r
m
, r = [x[, (10.43)
where
˜
P
lm
is a polynomial of degree l −m given by
˜
P
lm
(x) = (1 −x
2
)
−m/2
P
lm
(x) =
d
l+m
dx
l+m
(1 −x
2
)
l
. (10.44)
In particular, Y
lm
are smooth away from the origin and by construction they
satisfy
−∆Y
lm
=
l(l + 1)
r
2
Y
lm
. (10.45)
184 10. Oneparticle Schr¨odinger operators
10.4. The eigenvalues of the hydrogen atom
Now we want to use the considerations from the previous section to decom
pose the Hamiltonian of the hydrogen atom. In fact, we can even admit any
spherically symmetric potential V (x) = V ([x[) with
V (r) ∈ L
∞
∞
(R) +L
2
((0, ∞), r
2
dr). (10.46)
The important observation is that the spaces
H
lm
= ¦ψ(x) = R(r)Y
lm
(θ, ϕ)[R(r) ∈ L
2
((0, ∞), r
2
dr)¦ (10.47)
reduce our operator H = H
0
+V . Hence
H = H
0
+V =
l,m
˜
H
l
, (10.48)
where
˜
H
l
R(r) = ˜ τ
l
R(r), ˜ τ
l
= −
1
r
2
d
dr
r
2
d
dr
+
l(l + 1)
r
2
+V (r)
D(H
l
) ⊆ L
2
((0, ∞), r
2
dr). (10.49)
Using the unitary transformation
L
2
((0, ∞), r
2
dr) → L
2
((0, ∞)), R(r) → u(r) = rR(r), (10.50)
our operator transforms to
A
l
f = τ
l
f, τ
l
= −
d
2
dr
2
+
l(l + 1)
r
2
+V (r)
D(A
l
) ⊆ L
2
((0, ∞)). (10.51)
It remains to investigate this operator.
Theorem 10.8. The domain of the operator A
l
is given by
D(A
l
) = ¦f ∈ L
2
(I)[ f, f
∈ AC(I), τf ∈ L
2
(I),
lim
r→0
(f(r) −rf
(r)) = 0 if l = 0¦,
(10.52)
where I = (0, ∞). Moreover, σ
ess
(A
l
) = [0, ∞).
Proof. By construction of A
l
we know that it is selfadjoint and satisﬁes
σ
ess
(A
l
) = [0, ∞). Hence it remains to compute the domain. We know at
least D(A
l
) ⊆ D(τ) and since D(H) = D(H
0
) it suﬃces to consider the case
V = 0. In this case the solutions of −u
(r) +
l(l+1)
r
2
u(r) = 0 are given by
u(r) = αr
l+1
+ βr
−l
. Thus we are in the l.p. case at ∞ for any l ∈ N
0
.
However, at 0 we are in the l.p. case only if l > 0, that is, we need an
additional boundary condition at 0 if l = 0. Since we need R(r) =
u(r)
r
to
be bounded (such that (10.23) is in the domain of H
0
), we have to take the
boundary condition generated by u(r) = r.
10.4. The eigenvalues of the hydrogen atom 185
Finally let us turn to some explicit choices for V , where the correspond
ing diﬀerential equation can be explicitly solved. The simplest case is V = 0
in this case the solutions of
−u
(r) +
l(l + 1)
r
2
u(r) = zu(r) (10.53)
are given by the spherical Bessel respectively spherical Neumann func
tions
u(r) = αj
l
(
√
zr) +β n
l
(
√
zr), (10.54)
where
j
l
(r) = (−r)
l
1
r
d
dr
l
sin(r)
r
. (10.55)
In particular,
u
a
(z, r) = j
l
(
√
zr) and u
b
(z, r) = j
l
(
√
zr) + in
l
(
√
zr) (10.56)
are the functions which are square integrable and satisfy the boundary con
dition (if any) near a = 0 and b = ∞, respectively.
The second case is that of our Coulomb potential
V (r) = −
γ
r
, γ > 0, (10.57)
where we will try to compute the eigenvalues plus corresponding eigenfunc
tions. It turns out that they can be expressed in terms of the Laguerre
polynomials
L
j
(r) = e
r
d
j
dr
j
e
−r
r
j
(10.58)
and the associated Laguerre polynomials
L
k
j
(r) =
d
k
dr
k
L
j
(r). (10.59)
Note that L
k
j
is a polynomial of degree j −k.
Theorem 10.9. The eigenvalues of H
(1)
are explicitly given by
E
n
= −
γ
2(n + 1)
2
, n ∈ N
0
. (10.60)
An orthonormal basis for the corresponding eigenspace is given by
ψ
nlm
(x) = R
nl
(r)Y
lm
(x), (10.61)
where
R
nl
(r) =
γ
3
(n −l)!
2n
3
((n +l + 1)!)
3
γr
n + 1
l
e
−
γr
2(n+1)
L
2l+1
n+l+1
(
γr
n + 1
). (10.62)
In particular, the lowest eigenvalue E
0
= −
γ
2
4
is simple and the correspond
ing eigenfunction ψ
000
(x) =
γ
3
4
3
π
e
−γr/2
is positive.
186 10. Oneparticle Schr¨odinger operators
Proof. It is a straightforward calculation to check that R
nl
are indeed eigen
functions of A
l
corresponding to the eigenvalue −(
γ
2(n+1)
)
2
and for the norm
ing constants we refer to any book on special functions. The only problem
is to show that we have found all eigenvalues.
Since all eigenvalues are negative, we need to look at the equation
−u
(r) + (
l(l + 1)
r
2
−
γ
r
)u(r) = λu(r) (10.63)
for λ < 0. Introducing new variables x =
√
−λr and v(x) =
e
x
x
l+1
u(
x
√
−λ
)
this equation transforms into
xv
(x) + 2(l + 1 −x)v
(x) + 2nv(x) = 0, n =
γ
2
√
−λ
−(l + 1). (10.64)
Now let us search for a solution which can be expanded into a convergent
power series
v(x) =
∞
¸
j=0
v
j
x
j
, v
0
= 1. (10.65)
The corresponding u(r) is square integrable near 0 and satisﬁes the boundary
condition (if any). Thus we need to ﬁnd those values of λ for which it is
square integrable near +∞.
Substituting the ansatz (10.65) into our diﬀerential equation and com
paring powers of x gives the following recursion for the coeﬃcients
v
j+1
=
2(j −n)
(j + 1)(j + 2(l + 1))
v
j
(10.66)
and thus
v
j
=
1
j!
j−1
¸
k=0
2(k −n)
k + 2(l + 1)
. (10.67)
Now there are two cases to distinguish. If n ∈ N
0
, then v
j
= 0 for j > n
and v(x) is a polynomial. In this case u(r) is square integrable and hence an
eigenfunction corresponding to the eigenvalue λ
n
= −(
γ
2(n+l+1)
)
2
. Otherwise
we have v
j
≥
(2−ε)
j
j!
for j suﬃciently large. Hence by adding a polynomial
to v(x) we can get a function ˜ v(x) such that ˜ v
j
≥
(2−ε)
j
j!
for all j. But
then ˜ v(x) ≥ exp((2 − ε)x) and thus the corresponding u(r) is not square
integrable near −∞.
10.5. Nondegeneracy of the ground state
The lowest eigenvalue (below the essential spectrum) of a Schr¨odinger oper
ator is called ground state. Since the laws of physics state that a quantum
system will transfer energy to its surroundings (e.g., an atom emits radia
tion) until it eventually reaches its ground state, this state is in some sense
10.5. Nondegeneracy of the ground state 187
the most important state. We have seen that the hydrogen atom has a
nondegenerate (simple) ground state with a corresponding positive eigen
function. In particular, the hydrogen atom is stable in the sense that there
is a lowest possible energy. This is quite surprising since the corresponding
classical mechanical system is not, the electron could fall into the nucleus!
Our aim in this section is to show that the ground state is simple with a
corresponding positive eigenfunction. Note that it suﬃces to show that any
ground state eigenfunction is positive since nondegeneracy then follows for
free: two positive functions cannot be orthogonal.
To set the stage let us introduce some notation. Let H = L
2
(R
n
). We call
f ∈ L
2
(R
n
) positive if f ≥ 0 a.e. and f = 0. We call f strictly positive if
f > 0 a.e.. A bounded operator A is called positivity preserving if f ≥ 0
implies Af ≥ 0 and positivity improving if f ≥ 0 implies Af > 0. Clearly
A is positivity preserving (improving) if and only if 'f, Ag` ≥ 0 (> 0) for
f, g ≥ 0.
Example. Multiplication by a positive function is positivity preserving (but
not improving). Convolution with a strictly positive function is positivity
improving.
We ﬁrst show that positivity improving operators have positive eigen
functions.
Theorem 10.10. Suppose A ∈ L(L
2
(R
n
)) is a selfadjoint, positivity im
proving and real (i.e., it maps real functions to real functions) operator. If
A is an eigenvalue, then it is simple and the corresponding eigenfunction
is strictly positive.
Proof. Let ψ be an eigenfunction, then it is no restriction to assume that ψ
is real (since A is real both real and imaginary part of ψ are eigenfunctions
as well). We assume ψ = 1 and denote by ψ
±
=
f±f
2
the positive and
negative parts of ψ. Then by [Aψ[ = [Aψ
+
− Aψ
−
[ ≤ Aψ
+
+ Aψ
−
= A[ψ[
we have
A = 'ψ, Aψ` ≤ '[ψ[, [Aψ[` ≤ '[ψ[, A[ψ[` ≤ A, (10.68)
that is, 'ψ, Aψ` = '[ψ[, A[ψ[` and thus
'ψ
+
, Aψ
−
` =
1
4
('[ψ[, A[ψ[` −'ψ, Aψ`) = 0. (10.69)
Consequently ψ
−
= 0 or ψ
+
= 0 since otherwise Aψ
−
> 0 and hence also
'ψ
+
, Aψ
−
` > 0. Without restriction ψ = ψ
+
≥ 0 and since A is positivity
increasing we even have ψ = A
−1
Aψ > 0.
So we need a positivity improving operator. By (7.42) and (7.43) both
E
−tH
0
, t > 0 and R
λ
(H
0
), λ < 0 are since they are given by convolution
188 10. Oneparticle Schr¨odinger operators
with a strictly positive function. Our hope is that this property carries over
to H = H
0
+V .
Theorem 10.11. Suppose H = H
0
+ V is selfadjoint and bounded from
below with C
∞
c
(R
n
) as a core. If E
0
= min σ(H) is an eigenvalue, it is
simple and the corresponding eigenfunction is strictly positive.
Proof. We ﬁrst show that e
−tH
, t > 0, is positivity preserving. If we set
V
n
= V χ
{x V (x)≤n}
, then V
n
is bounded and H
n
= H
0
+ V
n
is positivity
preserving by the Trotter product formula since both e
−tH
0
and e
−tV
are.
Moreover, we have H
n
ψ → Hψ for ψ ∈ C
∞
c
(R
n
) (note that necessarily
V ∈ L
2
loc
) and hence H
n
sr
→ H in strong resolvent sense by Lemma 6.28.
Hence e
−tHn
s
→ e
−tH
by Theorem 6.23, which shows that e
−tH
is at least
positivity preserving (since 0 cannot be an eigenvalue of e
−tH
it cannot map
a positive function to 0).
Next I claim that for ψ positive the closed set
N(ψ) = ¦ϕ ∈ L
2
(R
n
) [ ϕ ≥ 0, 'ϕ, e
−sH
ψ` = 0∀s ≥ 0¦ (10.70)
is just ¦0¦. If ϕ ∈ N(ψ) we have by e
−sH
ψ ≥ 0 that ϕe
−sH
ψ = 0. Hence
e
tVn
ϕe
−sH
ψ = 0, that is e
tVn
ϕ ∈ N(ψ). In other words, both e
tVn
and e
−tH
leave N(ψ) invariant and invoking again Trotter’s formula the same is true
for
e
−t(H−Vn)
= slim
k→∞
e
−
t
k
H
e
t
k
Vn
k
. (10.71)
Since e
−t(H−Vn)
s
→ e
−tH
0
we ﬁnally obtain that e
−tH
0
leaves N(ψ) invariant,
but this operator is positivity increasing and thus N(ψ) = ¦0¦.
Now it remains to use (7.41) which shows
'ϕ, R
H
(λ)ψ` =
∞
0
e
λt
'ϕ, e
−tH
ψ`dt > 0, λ < E
0
, (10.72)
for ϕ, ψ positive. So R
H
(λ) is positivity increasing for λ < E
0
.
If ψ is an eigenfunction of H corresponding to E
0
it is an eigenfunction
of R
H
(λ) corresponding to
1
E
0
−λ
and the claim follows since R
H
(λ) =
1
E
0
−λ
.
Chapter 11
Atomic Schr¨odinger
operators
11.1. Selfadjointness
In this section we want to have a look at the Hamiltonian corresponding to
more than one interacting particle. It is given by
H = −
N
¸
j=1
∆
j
+
N
¸
j<k
V
j,k
(x
j
−x
k
). (11.1)
We ﬁrst consider the case of two particles, which will give us a feeling
for how the many particle case diﬀers from the one particle case and how
the diﬃculties can be overcome.
We denote the coordinates corresponding to the ﬁrst particle by x
1
=
(x
1,1
, x
1,2
, x
1,3
) and those corresponding to the second particle by x
2
=
(x
2,1
, x
2,2
, x
2,3
). If we assume that the interaction is again of the Coulomb
type, the Hamiltonian is given by
H = −∆
1
−∆
2
−
γ
[x
1
−x
2
[
, D(H) = H
2
(R
6
). (11.2)
Since Theorem 10.2 does not allow singularities for n ≥ 3, it does not tell
us whether H is selfadjoint or not. Let
(y
1
, y
2
) =
1
√
2
I I
−I I
(x
1
, x
2
), (11.3)
then H reads in this new coordinates
H = (−∆
1
) + (−∆
2
−
γ/
√
2
[y
2
[
). (11.4)
189
190 11. Atomic Schr¨odinger operators
In particular, it is the sum of a free particle plus a particle in an external
Coulomb ﬁeld. From a physics point of view, the ﬁrst part corresponds to
the center of mass motion and the second part to the relative motion.
Using that γ/(
√
2[y
2
[) has (−∆
2
)bound 0 in L
2
(R
3
) it is not hard to
see that the same is true for the (−∆
1
−∆
2
)bound in L
2
(R
6
) (details will
follow in the next section). In particular, H is selfadjoint and semibounded
for any γ ∈ R. Moreover, you might suspect that γ/(
√
2[y
2
[) is relatively
compact with respect to −∆
1
−∆
2
in L
2
(R
6
) since it is with respect to −∆
2
in L
2
(R
6
). However, this is not true! This is due to the fact that γ/(
√
2[y
2
[)
does not vanish as [y[ → ∞.
Let us look at this problem from the physical view point. If λ ∈ σ
ess
(H),
this means that the movement of the whole system is somehow unbounded.
There are two possibilities for this.
Firstly, both particles are far away from each other (such that we can
neglect the interaction) and the energy corresponds to the sum of the kinetic
energies of both particles. Since both can be arbitrarily small (but positive),
we expect [0, ∞) ⊆ σ
ess
(H).
Secondly, both particles remain close to each other and move together.
In the last coordinates this corresponds to a bound state of the second
operator. Hence we expect [λ
0
, ∞) ⊆ σ
ess
(H), where λ
0
= −γ
2
/8 is the
smallest eigenvalue of the second operator if the forces are attracting (γ ≥ 0)
and λ
0
= 0 if they are repelling (γ ≤ 0).
It is not hard to translate this intuitive ideas into a rigorous proof.
Let ψ
1
(y
1
) be a Weyl sequence corresponding to λ ∈ [0, ∞) for −∆
1
and
ψ
2
(y
2
) be a Weyl sequence corresponding to λ
0
for −∆
2
−γ/(
√
2[y
2
[). Then,
ψ
1
(y
1
)ψ
2
(y
2
) is a Weyl sequence corresponding to λ + λ
0
for H and thus
[λ
0
, ∞) ⊆ σ
ess
(H). Conversely, we have −∆
1
≥ 0 respectively −∆
2
−
γ/(
√
2[y
2
[) ≥ λ
0
and hence H ≥ λ
0
. Thus we obtain
σ(H) = σ
ess
(H) = [λ
0
, ∞), λ
0
=
−γ
2
/8, γ ≥ 0
0, γ ≤ 0
. (11.5)
Clearly, the physically relevant information is the spectrum of the operator
−∆
2
−γ/(
√
2[y
2
[) which is hidden by the spectrum of −∆
1
. Hence, in order
to reveal the physics, one ﬁrst has to remove the center of mass motion.
To avoid clumsy notation, we will restrict ourselves to the case of one
atom with N electrons whose nucleus is ﬁxed at the origin. In particular,
this implies that we do not have to deal with the center of mass motion
11.2. The HVZ theorem 191
encountered in our example above. The Hamiltonian is given by
H
(N)
= −
N
¸
j=1
∆
j
−
N
¸
j=1
V
ne
(x
j
) +
N
¸
j=1
N
¸
j<k
V
ee
(x
j
−x
k
),
D(H
(N)
) = H
2
(R
3N
), (11.6)
where V
ne
describes the interaction of one electron with the nucleus and V
ee
describes the interaction of two electrons. Explicitly we have
V
j
(x) =
γ
j
[x[
, γ
j
> 0, j = ne, ee. (11.7)
We ﬁrst need to establish selfadjointness of H
(N)
. This will follow from
Kato’s theorem.
Theorem 11.1 (Kato). Let V
k
∈ L
∞
∞
(R
d
) + L
2
(R
d
), d ≤ 3, be realvalued
and let V
k
(y
(k)
) be the multiplication operator in L
2
(R
n
), n = Nd, obtained
by letting y
(k)
be the ﬁrst d coordinates of a unitary transform of R
n
. Then
V
k
is H
0
bounded with H
0
bound 0. In particular,
H = H
0
+
¸
k
V
k
(y
(k)
), D(H) = H
2
(R
n
), (11.8)
is selfadjoint and C
∞
0
(R
n
) is a core.
Proof. It suﬃces to consider one k. After a unitary transform of R
n
we can
assume y
(1)
= (x
1
, . . . , x
d
) since such transformations leave both the scalar
product of L
2
(R
n
) and H
0
invariant. Now let ψ ∈ o(R
n
), then
V
k
ψ
2
≤ a
2
R
n
[∆
1
ψ(x)[
2
d
n
x +b
2
R
n
[ψ(x)[
2
d
n
x, (11.9)
where ∆
1
=
¸
d
j=1
∂
2
/∂
2
x
j
, by our previous lemma. Hence we obtain
V
k
ψ
2
≤ a
2
R
n
[
d
¸
j=1
p
2
j
ˆ
ψ(p)[
2
d
n
p +b
2
ψ
2
≤ a
2
R
n
[
n
¸
j=1
p
2
j
ˆ
ψ(p)[
2
d
n
p +b
2
ψ
2
= a
2
H
0
ψ
2
+b
2
ψ
2
, (11.10)
which implies that V
k
is relatively bounded with bound 0.
11.2. The HVZ theorem
The considerations of the beginning of this section show that it is not so
easy to determine the essential spectrum of H
(N)
since the potential does
not decay in all directions as [x[ → ∞. However, there is still something we
192 11. Atomic Schr¨odinger operators
can do. Denote the inﬁmum of the spectrum of H
(N)
by λ
N
. Then, let us
split the system into H
(N−1)
plus a single electron. If the single electron is
far away from the remaining system such that there is little interaction, the
energy should be the sum of the kinetic energy of the single electron and
the energy of the remaining system. Hence arguing as in the two electron
example of the previous section we expect
Theorem 11.2 (HVZ). Let H
(N)
be the selfadjoint operator given in (11.6).
Then H
(N)
is bounded from below and
σ
ess
(H
(N)
) = [λ
N−1
, ∞), (11.11)
where λ
N−1
= min σ(H
(N−1)
) < 0.
In particular, the ionization energy (i.e., the energy needed to remove
one electron from the atom in its ground state) of an atom with N electrons
is given by λ
N
−λ
N−1
.
Our goal for the rest of this section is to prove this result which is due to
Zhislin, van Winter and Hunziker and known as HVZ theorem. In fact there
is a version which holds for general Nbody systems. The proof is similar
but involves some additional notation.
The idea of proof is the following. To prove [λ
N−1
, ∞) ⊆ σ
ess
(H
(N)
)
we choose Weyl sequences for H
(N−1)
and −∆
N
and proceed according to
our intuitive picture from above. To prove σ
ess
(H
(N)
) ⊆ [λ
N−1
, ∞) we will
localize H
(N)
on sets where either one electron is far away from the others
or all electrons are far away from the nucleus. Since the error turns out
relatively compact, it remains to consider the inﬁmum of the spectra of
these operators. For all cases where one electron is far away it is λ
N−1
and
for the case where all electrons are far away from the nucleus it is 0 (since
the electrons repel each other).
We begin with the ﬁrst inclusion. Let ψ
N−1
(x
1
, . . . , x
N−1
) ∈ H
2
(R
3(N−1)
)
such that ψ
N−1
 = 1, (H
(N−1)
− λ
N−1
)ψ
N−1
 ≤ ε and ψ
1
∈ H
2
(R
3
)
such that ψ
1
 = 1, (−∆
N
− λ)ψ
1
 ≤ ε for some λ ≥ 0. Now consider
ψ
r
(x
1
, . . . , x
N
) = ψ
N−1
(x
1
, . . . , x
N−1
)ψ
1
r
(x
N
), ψ
1
r
(x
N
) = ψ
1
(x
N
−r), then
(H
(N)
−λ −λ
N−1
)ψ
r
 ≤ (H
(N−1)
−λ
N−1
)ψ
N−1
ψ
1
r

+ψ
N−1
(−∆
N
−λ)ψ
1
r

+(V
N
−
N−1
¸
j=1
V
N,j
)ψ
r
, (11.12)
where V
N
= V
ne
(x
N
) and V
N,j
= V
ee
(x
N
−x
j
). Since (V
N
−
¸
N−1
j=1
V
N,j
)ψ
N−1
∈
L
2
(R
3N
) and [ψ
1
r
[ → 0 pointwise as [r[ → ∞ (by Lemma 10.1), the third
11.2. The HVZ theorem 193
term can be made smaller than ε by choosing [r[ large (dominated conver
gence). In summary,
(H
(N)
−λ −λ
N−1
)ψ
r
 ≤ 3ε (11.13)
proving [λ
N−1
, ∞) ⊆ σ
ess
(H
(N)
).
The second inclusion is more involved. We begin with a localization
formula, which can be veriﬁed by a straightforward computation
Lemma 11.3 (IMS localization formula). Suppose φ
j
∈ C
∞
(R
n
), 0 ≤ j ≤
N, is such that
N
¸
j=0
φ
j
(x)
2
= 1, x ∈ R
n
, (11.14)
then
∆ψ =
N
¸
j=0
φ
j
∆φ
j
ψ −[∂φ
j
[
2
ψ, ψ ∈ H
2
(R
n
). (11.15)
Abbreviate B = ¦x ∈ R
3N
[[x[ ≥ 1¦. Now we will choose φ
j
, 1 ≤ j ≤ N,
in such a way that x ∈ supp(φ
j
) ∩ B implies that the jth particle is far
away from all the others and from the nucleus. Similarly, we will choose φ
0
in such a way that x ∈ supp(φ
0
) ∩ B implies that all particle are far away
from the nucleus.
Lemma 11.4. There exists functions φ
j
∈ C
∞
(R
n
, [0, 1]), 0 ≤ j ≤ N, is
such that (11.14) holds,
supp(φ
j
) ∩ B ⊆ ¦x ∈ B[ [x
j
−x
[ ≥ C[x[ for all = j, and [x
j
[ ≥ C[x[¦,
supp(φ
0
) ∩ B ⊆ ¦x ∈ B[ [x
[ ≥ C[x[ for all ¦ (11.16)
for some C ∈ [0, 1], and [∂φ
j
(x)[ → 0 as [x[ → ∞.
Proof. Consider the sets
U
n
j
= ¦x ∈ S
3N−1
[ [x
j
−x
[ > n
−1
for all = j, and [x
j
[ > n
−1
¦,
U
n
0
= ¦x ∈ S
3N−1
[ [x
[ > n
−1
for all ¦. (11.17)
We claim that
∞
¸
n=1
N
¸
j=0
U
n
j
= S
3N−1
. (11.18)
Indeed, suppose there is an x ∈ S
3N−1
which is not an element of this union.
Then x ∈ U
n
0
for all n implies 0 = [x
j
[ for some j, say j = 1. Next, since
x ∈ U
n
1
for all n implies 0 = [x
j
− x
1
[ = [x
j
[ for some j > 1, say j = 2.
194 11. Atomic Schr¨odinger operators
Proceeding like this we end up with x = 0, a contradiction. By compactness
of S
3N−1
we even have
N
¸
j=0
U
n
j
= S
3N−1
(11.19)
for n suﬃciently large. It is wellknown that there is a partition of unity
˜
φ
j
(x) subordinate to this cover. Extend
˜
φ
j
(x) to a smooth function from
R
3N
`¦0¦ to [0, 1] by
˜
φ
j
(λx) =
˜
φ
j
(x), x ∈ S
3N−1
, λ > 0, (11.20)
and pick a function
˜
φ ∈ C
∞
(R
3N
, [0, 1]) with support inside the unit ball
which is 1 in a neighborhood of the origin. Then
φ
j
=
˜
φ + (1 −
˜
φ)
˜
φ
j
¸
N
=0
˜
φ + (1 −
˜
φ)
˜
φ
(11.21)
are the desired functions. The gradient tends to zero since φ
j
(λx) = φ
j
(x)
for λ ≥ 1 and [x[ ≥ 1 which implies (∂φ
j
)(λx) = λ
−1
(∂φ
j
)(x).
By our localization formula we have
H
(N)
=
N
¸
j=0
φ
j
H
(N,j)
φ
j
+K, K =
N
¸
j=0
φ
2
j
V
(N,j)
+[∂φ
j
[
2
, (11.22)
where
H
(N,j)
= −
N
¸
=1
∆
−
N
¸
=j
V
+
N
¸
k<, k,=j
V
k,
, H
(N,0)
= −
N
¸
=1
∆
+
N
¸
k<
V
k,
V
(N,j)
= V
j
+
N
¸
=j
V
j,
, V
(N,0)
=
N
¸
=1
V
(11.23)
To show that our choice of the functions φ
j
implies that K is relatively
compact with respect to H we need the following
Lemma 11.5. Let V be H
0
bounded with H
0
bound 0 and suppose that
χ
{xx≥R}
V R
H
0
(z) → 0 as R → ∞. Then V is relatively compact with
respect to H
0
.
Proof. Let ψ
n
converge to 0 weakly. Note that ψ
n
 ≤ M for some
M > 0. It suﬃces to show that V R
H
0
(z)ψ
n
 converges to 0. Choose
φ ∈ C
∞
0
(R
n
, [0, 1]) such that it is one for [x[ ≤ R. Then
V R
H
0
(z)ψ
n
 ≤ (1 −φ)V R
H
0
(z)ψ
n
 +V φR
H
0
(z)ψ
n

≤ (1 −φ)V R
H
0
(z)
∞
ψ
n
 +
aH
0
φR
H
0
(z)ψ
n
 +bφR
H
0
(z)ψ
n
. (11.24)
11.2. The HVZ theorem 195
By assumption, the ﬁrst term can be made smaller than ε by choosing R
large. Next, the same is true for the second term choosing a small. Finally,
the last term can also be made smaller than ε by choosing n large since φ
is H
0
compact.
The terms [∂φ
j
[
2
are bounded and vanish at ∞, hence they are H
0
compact by Lemma 7.10. The terms φ
j
V
(N,j)
are relatively compact by
the lemma and hence K is relatively compact with respect to H
0
. By
Lemma 6.22, K is also relatively compact with respect to H
(N)
since V
(N)
is relatively bounded with respect to H
0
.
In particular H
(N)
− K is selfadjoint on H
2
(R
3N
) and σ
ess
(H
(N)
) =
σ
ess
(H
(N)
− K). Since the operators H
(N,j)
, 1 ≤ j ≤ N, are all of the
form H
(N−1)
plus one particle which does not interact with the others and
the nucleus, we have H
(N,j)
− λ
N−1
≥ 0, 1 ≤ j ≤ N. Moreover, we have
H
(0)
≥ 0 since V
j,k
≥ 0 and hence
'ψ, (H
(N)
−K −λ
N−1
)ψ` =
N
¸
j=0
'φ
j
ψ, (H
(N,j)
−λ
N−1
)φ
j
ψ` ≥ 0. (11.25)
Thus we obtain the remaining inclusion
σ
ess
(H
(N)
) = σ
ess
(H
(N)
−K) ⊆ σ(H
(N)
−K) ⊆ [λ
N−1
, ∞) (11.26)
which ﬁnishes the proof of the HVZ theorem.
Note that the same proof works if we add additional nuclei at ﬁxed
locations. That is, we can also treat molecules if we assume that the nuclei
are ﬁxed in space.
Finally, let us consider the example of Helium like atoms (N = 2). By
the HVZ theorem and the considerations of the previous section we have
σ
ess
(H
(2)
) = [−
γ
2
ne
4
, ∞). (11.27)
Moreover, if γ
ee
= 0 (no electron interaction), we can take products of one
particle eigenfunctions to show that
−γ
2
ne
1
4n
2
+
1
4m
2
∈ σ
p
(H
(2)
(γ
ee
= 0)), n, m ∈ N. (11.28)
In particular, there are eigenvalues embedded in the essential spectrum in
this case. Moreover, since the electron interaction term is positive, we see
H
(2)
≥ −
γ
2
ne
2
. (11.29)
Note that there can be no positive eigenvalues by the virial theorem. This
even holds for arbitrary N,
σ
p
(H
(N)
) ⊂ (−∞, 0). (11.30)
Chapter 12
Scattering theory
12.1. Abstract theory
In physical measurements one often has the following situation. A particle
is shot into a region where it interacts with some forces and then leaves
the region again. Outside this region the forces are negligible and hence the
time evolution should be asymptotically free. Hence one expects asymptotic
states ψ
±
(t) = exp(−itH
0
)ψ
±
(0) to exist such that
ψ(t) −ψ
±
(t) → 0 as t → ±∞. (12.1)
ψ(t)
ψ
−
(t)
ψ
+
(t)
¡
¡
¡
¡
¡
¡
¡
¡
¡
¡
¡
¡
¡
¡
¡!
$
$
$
$
$
$
$
$
$
$
$
$
$
$
$
$
$$X
Rewriting this condition we see
0 = lim
t→±∞
e
−itH
ψ(0) −e
−itH
0
ψ
±
(0) = lim
t→±∞
ψ(0) −e
itH
e
−itH
0
ψ
±
(0)
(12.2)
and motivated by this we deﬁne the wave operators by
D(Ω
±
) = ¦ψ ∈ H[∃ lim
t→±∞
e
itH
e
−itH
0
ψ¦
Ω
±
ψ = lim
t→±∞
e
itH
e
−itH
0
ψ
. (12.3)
197
198 12. Scattering theory
The set D(Ω
±
) is the set of all incoming/outgoing asymptotic states ψ
±
and
Ran(Ω
±
) is the set of all states which have an incoming/outgoing asymptotic
state. If a state ψ has both, that is, ψ ∈ Ran(Ω
+
) ∩ Ran(Ω
−
), it is called a
scattering state.
By construction we have
Ω
±
ψ = lim
t→±∞
e
itH
e
−itH
0
ψ = lim
t→±∞
ψ = ψ (12.4)
and it is not hard to see that D(Ω
±
) is closed. Moreover, interchanging the
roles of H
0
and H amounts to replacing Ω
±
by Ω
−1
±
and hence Ran(Ω
±
) is
also closed. In summary,
Lemma 12.1. The sets D(Ω
±
) and Ran(Ω
±
) are closed and Ω
±
: D(Ω
±
) →
Ran(Ω
±
) is unitary.
Next, observe that
lim
t→±∞
e
itH
e
−itH
0
(e
−isH
0
ψ) = lim
t→±∞
e
−isH
(e
i(t+s)H
e
−i(t+s)H
0
ψ) (12.5)
and hence
Ω
±
e
−itH
0
ψ = e
−itH
Ω
±
ψ, ψ ∈ D(Ω
±
). (12.6)
In addition, D(Ω
±
) is invariant under exp(−itH
0
) and Ran(Ω
±
) is invariant
under exp(−itH). Moreover, if ψ ∈ D(Ω
±
)
⊥
then
'ϕ, exp(−itH
0
)ψ` = 'exp(itH
0
)ϕ, ψ` = 0, ϕ ∈ D(Ω
±
). (12.7)
Hence D(Ω
±
)
⊥
is invariant under exp(−itH
0
) and Ran(Ω
±
)
⊥
is invariant
under exp(−itH). Consequently, D(Ω
±
) reduces exp(−itH
0
) and Ran(Ω
±
)
reduces exp(−itH). Moreover, diﬀerentiating (12.6) with respect to t we
obtain from Theorem 5.1 the intertwining property of the wave operators.
Theorem 12.2. The subspaces D(Ω
±
) respectively Ran(Ω
±
) reduce H
0
re
spectively H and the operators restricted to these subspaces are unitarily
equivalent
Ω
±
H
0
ψ = HΩ
±
ψ, ψ ∈ D(Ω
±
) ∩ D(H
0
). (12.8)
It is interesting to know the correspondence between incoming and out
going states. Hence we deﬁne the scattering operator
S = Ω
−1
+
Ω
−
, D(S) = ¦ψ ∈ D(Ω
−
)[Ω
−
ψ ∈ Ran(Ω
+
)¦. (12.9)
Note that we have D(S) = D(Ω
−
) if and only if Ran(Ω
−
) ⊆ Ran(Ω
+
) and
Ran(S) = D(Ω
+
) if and only if Ran(Ω
+
) ⊆ Ran(Ω
−
). Moreover, S is unitary
from D(S) onto Ran(S) and we have
H
0
Sψ = SH
0
ψ, D(H
0
) ∩ D(S). (12.10)
However, note that this whole theory is meaningless until we can show
that D(Ω
±
) are nontrivial. We ﬁrst show a criterion due to Cook.
12.1. Abstract theory 199
Lemma 12.3 (Cook). Suppose D(H) ⊆ D(H
0
). If
∞
0
(H −H
0
) exp(∓itH
0
)ψdt < ∞, ψ ∈ D(H
0
), (12.11)
then ψ ∈ D(Ω
±
), respectively. Moreover, we even have
(Ω
±
−I)ψ ≤
∞
0
(H −H
0
) exp(∓itH
0
)ψdt (12.12)
in this case.
Proof. The result follows from
e
itH
e
−itH
0
ψ = ψ + i
t
0
exp(isH)(H −H
0
) exp(−isH
0
)ψds (12.13)
which holds for ψ ∈ D(H
0
).
As a simple consequence we obtain the following result for Schr¨odinger
operators in R
3
Theorem 12.4. Suppose H
0
is the free Schr¨odinger operator and H =
H
0
+V with V ∈ L
2
(R
3
), then the wave operators exist and D(Ω
±
) = H.
Proof. Since we want to use Cook’s lemma, we need to estimate
V ψ(s)
2
=
R
3
[V (x)ψ(s, x)[
2
dx, ψ(s) = exp(isH
0
)ψ, (12.14)
for given ψ ∈ D(H
0
). Invoking (7.34) we get
V ψ(s) ≤ ψ(s)
∞
V  ≤
1
(4πs)
3/2
ψ
1
V , s > 0, (12.15)
at least for ψ ∈ L
1
(R
3
). Moreover, this implies
∞
1
V ψ(s)ds ≤
1
4π
3/2
ψ
1
V  (12.16)
and thus any such ψ is in D(Ω
+
). Since such functions are dense, we obtain
D(Ω
+
) = H. Similarly for Ω
−
.
By the intertwining property ψ is an eigenfunction of H
0
if and only
if it is an eigenfunction of H. Hence for ψ ∈ H
pp
(H
0
) it is easy to check
whether it is in D(Ω
±
) or not and only the continuous subspace is of interest.
We will say that the wave operators exist if all elements of H
ac
(H
0
) are
asymptotic states, that is,
H
ac
(H
0
) ⊆ D(Ω
±
) (12.17)
and that they are complete if, in addition, all elements of H
ac
(H) are
scattering states, that is,
H
ac
(H) ⊆ Ran(Ω
±
). (12.18)
200 12. Scattering theory
If we even have
H
c
(H) ⊆ Ran(Ω
±
), (12.19)
they are called asymptotically complete. We will be mainly interested in
the case where H
0
is the free Schr¨odinger operator and hence H
ac
(H
0
) = H.
In this later case the wave operators exist if D(Ω
±
) = H, they are complete if
H
ac
(H) = Ran(Ω
±
), and asymptotically complete if H
c
(H) = Ran(Ω
±
). In
particular asymptotic completeness implies H
sc
(H) = ∅ since H restricted
to Ran(Ω
±
) is unitarily equivalent to H
0
.
12.2. Incoming and outgoing states
In the remaining sections we want to apply this theory to Schr¨odinger op
erators. Our ﬁrst goal is to give a precise meaning to some terms in the
intuitive picture of scattering theory introduced in the previous section.
This physical picture suggests that we should be able to decompose
ψ ∈ H into an incoming and an outgoing part. But how should incom
ing respectively outgoing be deﬁned for ψ ∈ H? Well incoming (outgoing)
means that the expectation of x
2
should decrease (increase). Set x(t)
2
=
exp(iH
0
t)x
2
exp(−iH
0
t), then, abbreviating ψ(t) = e
−itH
0
ψ,
d
dt
E
ψ
(x(t)
2
) = 'ψ(t), i[H
0
, x
2
]ψ(t)` = 4'ψ(t), Dψ(t)`, ψ ∈ o(R
n
),
(12.20)
where D is the dilation operator introduced in (10.11). Hence it is natural
to consider ψ ∈ Ran(P
±
),
P
±
= P
D
((0, ±∞)), (12.21)
as outgoing respectively incoming states. If we project a state in Ran(P
±
)
to energies in the interval (a
2
, b
2
), we expect that it cannot be found in a
ball of radius proportional to a[t[ as t → ±∞ (a is the minimal velocity of
the particle, since we have assumed the mass to be two). In fact, we will
show below that the tail decays faster then any inverse power of [t[.
We ﬁrst collect some properties of D which will be needed later on. Note
TD = −DT (12.22)
and hence Tf(D) = f(−D)T. To say more we will look for a transformation
which maps D to a multiplication operator.
Since the dilation group acts on [x[ only, it seems reasonable to switch
to polar coordinates x = rω, (t, ω) ∈ R
+
S
n−1
. Since U(s) essentially
transforms r into r exp(s) we will replace r by ρ = ln(r). In these coordinates
we have
U(s)ψ(e
ρ
ω) = e
−ns/2
ψ(e
(ρ−s)
ω) (12.23)
12.2. Incoming and outgoing states 201
and hence U(s) corresponds to a shift of ρ (the constant in front is absorbed
by the volume element). Thus D corresponds to diﬀerentiation with respect
to this coordinate and all we have to do to make it a multiplication operator
is to take the Fourier transform with respect to ρ.
This leads us to the Mellin transform
´ : L
2
(R
n
) → L
2
(R S
n−1
)
ψ(rω) → (´ψ)(λ, ω) =
1
√
2π
∞
0
r
−iλ
ψ(rω)r
n
2
−1
dr
. (12.24)
By construction, ´ is unitary, that is,
R
S
n−1
[(´ψ)(λ, ω)[
2
dλd
n−1
ω =
R
+
S
n−1
[ψ(rω)[
2
r
n−1
drd
n−1
ω,
(12.25)
where d
n−1
ω is the normalized surface measure on S
n−1
. Moreover,
´
−1
U(s)´ = e
−isλ
(12.26)
and hence
´
−1
D´ = λ. (12.27)
From this it is straightforward to show that
σ(D) = σ
ac
(D) = R, σ
sc
(D) = σ
pp
(D) = ∅ (12.28)
and that o(R
n
) is a core for D. In particular we have P
+
+P
−
= I.
Using the Mellin transform we can now prove Perry’s estimate [11].
Lemma 12.5. Suppose f ∈ C
∞
c
(R) with supp(f) ⊂ (a
2
, b
2
) for some a, b >
0. For any R ∈ R, N ∈ N there is a constant C such that
χ
{x x<2at}
e
−itH
0
f(H
0
)P
D
((±R, ±∞)) ≤
C
(1 +[t[)
N
, ±t ≥ 0, (12.29)
respectively.
Proof. We prove only the + case, the remaining one being similar. Consider
ψ ∈ o(R
n
). Introducing
ψ(t, x) = e
−itH
0
f(H
0
)P
D
((R, ∞))ψ(x) = 'K
t,x
, TP
D
((R, ∞))ψ`
= 'K
t,x
, P
D
((−∞, −R))
ˆ
ψ`, (12.30)
where
K
t,x
(p) =
1
(2π)
n/2
e
i(tp
2
−px)
f(p
2
)
∗
, (12.31)
we see that it suﬃces to show
P
D
((−∞, −R))K
t,x

2
≤
const
(1 +[t[)
2N
, for [x[ < 2a[t[, t > 0. (12.32)
202 12. Scattering theory
Now we invoke the Mellin transform to estimate this norm
P
D
((−∞, −R))K
t,x

2
=
R
−∞
S
n−1
[(´K
t,x
)(λ, ω)[
2
dλd
n−1
ω. (12.33)
Since
(´K
t,x
)(λ, ω) =
1
(2π)
(n+1)/2
∞
0
˜
f(r)e
iα(r)
dr (12.34)
with
˜
f(r) = f(r
2
)
∗
r
n/2−1
∈ C
∞
c
((a
2
, b
2
)), α(r) = tr
2
+ rωx − λln(r). Esti
mating the derivative of α we see
α
(r) = 2tr +ωx −λ/r > 0, r ∈ (a, b), (12.35)
for λ ≤ −R and t > R(2εa)
−1
, where ε is the distance of a to the support
of
˜
f. Hence we can ﬁnd a constant such that
1
[α
(r)[
≤
const
1 +[λ[ +[t[
, r ∈ (a, b), (12.36)
and λ, t as above. Using this we can estimate the integral in (12.34)
∞
0
˜
f(r)
1
α
(r)
d
dr
e
iα(r)
dr
≤
const
1 +[λ[ +[t[
∞
0
˜
f
(r)e
iα(r)
dr
, (12.37)
(the last step uses integration by parts) for λ, t as above. By increasing the
constant we can even assume that it holds for t ≥ 0 and λ ≤ −R. Moreover,
by iterating the last estimate we see
[(´K
t,x
)(λ, ω)[ ≤
const
(1 +[λ[ +[t[)
N
(12.38)
for any N ∈ N and t ≥ 0 and λ ≤ −R. This ﬁnishes the proof.
Corollary 12.6. Suppose that f ∈ C
∞
c
((0, ∞)) and R ∈ R. Then the
operator P
D
((±R, ±∞))f(H
0
) exp(−itH
0
) converges strongly to 0 as t →
∓∞.
Proof. Abbreviating P
D
= P
D
((±R, ±∞)) and χ = χ
{x x<2at}
we have
P
D
f(H
0
)e
−itH
0
ψ ≤ χe
itH
0
f(H
0
)
∗
P
D
 ψ+f(H
0
)(I−χ)ψ. (12.39)
since A = A
∗
. Taking t → ∓∞ the ﬁrst term goes to zero by our lemma
and the second goes to zero since χψ → ψ.
12.3. Schr¨odinger operators with short range
potentials
By the RAGE theorem we know that for ψ ∈ H
c
, ψ(t) will eventually leave
every compact ball (at least on the average). Hence we expect that the
time evolution will asymptotically look like the free one for ψ ∈ H
c
if the
potential decays suﬃciently fast. In other words, we expect such potentials
to be asymptotically complete.
12.3. Schr¨odinger operators with short range potentials 203
Suppose V is relatively bounded with bound less than one. Introduce
h
1
(r) = V R
H
0
(z)χ
r
, h
2
(r) = χ
r
V R
H
0
(z), r ≥ 0, (12.40)
where
χ
r
= χ
{x x≥r}
. (12.41)
The potential V will be called short range if these quantities are integrable.
We ﬁrst note that it suﬃces to check this for h
1
or h
2
and for one z ∈ ρ(H
0
).
Lemma 12.7. The function h
1
is integrable if and only if h
2
is. Moreover,
h
j
integrable for one z
0
∈ ρ(H
0
) implies h
j
integrable for all z ∈ ρ(H
0
).
Proof. Pick φ ∈ C
∞
c
(R
n
, [0, 1]) such that φ(x) = 0 for 0 ≤ [x[ ≤ 1/2 and
φ(x) = 0 for 1 ≤ [x[. Then it is not hard to see that h
j
is integrable if and
only if
˜
h
j
is integrable, where
˜
h
1
(r) = V R
H
0
(z)φ
r
,
˜
h
2
(r) = φ
r
V R
H
0
(z), r ≥ 1, (12.42)
and φ
r
(x) = φ(x/r). Using
[R
H
0
(z), φ
r
] = −R
H
0
(z)[H
0
(z), φ
r
]R
H
0
(z)
= R
H
0
(z)(∆φ
r
+ (∂φ
r
)∂)R
H
0
(z) (12.43)
and ∆φ
r
= φ
r/2
∆φ
r
, ∆φ
r

∞
≤ ∆φ
∞
/r
2
respectively (∂φ
r
) = φ
r/2
(∂φ
r
),
∂φ
r

∞
≤ ∂φ
∞
/r
2
we see
[
˜
h
1
(r) −
˜
h
2
(r)[ ≤
c
r
˜
h
1
(r/2), r ≥ 1. (12.44)
Hence
˜
h
2
is integrable if
˜
h
1
is. Conversely,
˜
h
1
(r) ≤
˜
h
2
(r) +
c
r
˜
h
1
(r/2) ≤
˜
h
2
(r) +
c
r
˜
h
2
(r/2) +
2c
r
2
˜
h
1
(r/4) (12.45)
shows that
˜
h
2
is integrable if
˜
h
1
is.
Invoking the ﬁrst resolvent formula
φ
r
V R
H
0
(z) ≤ φ
r
V R
H
0
(z
0
)I −(z −z
0
)R
H
0
(z) (12.46)
ﬁnishes the proof.
As a ﬁrst consequence note
Lemma 12.8. If V is short range, then R
H
(z) −R
H
0
(z) is compact.
Proof. The operator R
H
(z)V (I−χ
r
)R
H
0
(z) is compact since (I−χ
r
)R
H
0
(z)
is by Lemma 7.10 and R
H
(z)V is bounded by Lemma 6.22. Moreover, by
our short range condition it converges in norm to
R
H
(z)V R
H
0
(z) = R
H
(z) −R
H
0
(z) (12.47)
as r → ∞ (at least for some subsequence).
204 12. Scattering theory
In particular, by Weyl’s theorem we have σ
ess
(H) = [0, ∞). Moreover,
V short range implies that H and H
0
look alike far outside.
Lemma 12.9. Suppose R
H
(z)−R
H
0
(z) is compact, then so is f(H)−f(H
0
)
for any f ∈ C
∞
(R) and
lim
r→∞
(f(H) −f(H
0
))χ
r
 = 0. (12.48)
Proof. The ﬁrst part is Lemma 6.20 and the second part follows from part
(ii) of Lemma 6.8 since χ
r
converges strongly to 0.
However, this is clearly not enough to prove asymptotic completeness
and we need a more careful analysis. The main ideas are due to Enß [4].
We begin by showing that the wave operators exist. By Cook’s criterion
(Lemma 12.3) we need to show that
V exp(∓itH
0
)ψ ≤ V R
H
0
(−1)(I −χ
2at
) exp(∓itH
0
)(H
0
+I)ψ
+V R
H
0
(−1)χ
2at
(H
0
+I)ψ (12.49)
is integrable for a dense set of vectors ψ. The second term is integrable by our
short range assumption. The same is true by Perry’s estimate (Lemma 12.5)
for the ﬁrst term if we choose ψ = f(H
0
)P
D
((±R, ±∞))ϕ. Since vectors of
this form are dense, we see that the wave operators exist,
D(Ω
±
) = H. (12.50)
Since H restricted to Ran(Ω
∗
±
) is unitarily equivalent to H
0
, we obtain
[0, ∞) = σ
ac
(H
0
) ⊆ σ
ac
(H). And by σ
ac
(H) ⊆ σ
ess
(H) = [0, ∞) we even
have σ
ac
(H) = [0, ∞).
To prove asymptotic completeness of the wave operators we will need
that (Ω
±
−I)f(H
0
)P
±
are compact.
Lemma 12.10. Let f ∈ C
∞
c
((0, ∞)) and suppose ψ
n
converges weakly to 0.
Then
lim
n→∞
(Ω
±
−I)f(H
0
)P
±
ψ
n
 = 0, (12.51)
that is, (Ω
±
−I)f(H
0
)P
±
is compact.
Proof. By (12.13) we see
R
H
(z)(Ω
±
−I)f(H
0
)P
±
ψ
n
 ≤
∞
0
R
H
(z)V exp(−isH
0
)f(H
0
)P
±
ψ
n
dt.
(12.52)
Since R
H
(z)V R
H
0
is compact we see that the integrand
R
H
(z)V exp(−isH
0
)f(H
0
)P
±
ψ
n
=
R
H
(z)V R
H
0
exp(−isH
0
)(H
0
+ 1)f(H
0
)P
±
ψ
n
(12.53)
12.3. Schr¨odinger operators with short range potentials 205
converges pointwise to 0. Moreover, arguing as in (12.49) the integrand
is bounded by an L
1
function depending only on ψ
n
. Thus R
H
(z)(Ω
±
−
I)f(H
0
)P
±
is compact by the dominated convergence theorem. Furthermore,
using the intertwining property we see that
(Ω
±
−I)
˜
f(H
0
)P
±
= R
H
(z)(Ω
±
−I)f(H
0
)P
±
−(R
H
(z) −R
H
0
(z))f(H
0
)P
±
(12.54)
is compact by Lemma 6.20, where
˜
f(λ) = (λ + 1)f(λ).
Now we have gathered enough information to tackle the problem of
asymptotic completeness.
We ﬁrst show that the singular continuous spectrum is absent. This
is not really necessary, but avoids the use of Ces`aro means in our main
argument.
Abbreviate P = P
sc
H
P
H
((a, b)), 0 < a < b. Since H restricted to
Ran(Ω
±
) is unitarily equivalent to H
0
(which has purely absolutely continu
ous spectrum), the singular part must live on Ran(Ω
±
)
⊥
, that is, P
sc
H
Ω
±
= 0.
Thus Pf(H
0
) = P(I −Ω
+
)f(H
0
)P
+
+P(I −Ω
−
)f(H
0
)P
−
is compact. Since
f(H) − f(H
0
) is compact, it follows that Pf(H) is also compact. Choos
ing f such that f(λ) = 1 for λ ∈ [a, b] we see that P = Pf(H) is com
pact and hence ﬁnite dimensional. In particular σ
sc
(H) ∩ (a, b) is a ﬁ
nite set. But a continuous measure cannot be supported on a ﬁnite set,
showing σ
sc
(H) ∩ (a, b) = ∅. Since 0 < a < b are arbitrary we even
have σ
sc
(H) ∩ (0, ∞) = ∅ and by σ
sc
(H) ⊆ σ
ess
(H) = [0, ∞) we obtain
σ
sc
(H) = ∅.
Observe that replacing P
sc
H
by P
pp
H
the same argument shows that all
nonzero eigenvalues are ﬁnite dimensional and cannot accumulate in (0, ∞).
In summary we have shown
Theorem 12.11. Suppose V is short range. Then
σ
ac
(H) = σ
ess
(H) = [0, ∞), σ
sc
(H) = ∅. (12.55)
All nonzero eigenvalues have ﬁnite multiplicity and cannot accumulate in
(0, ∞).
Now we come to the anticipated asymptotic completeness result of Enß.
Choose
ψ ∈ H
c
(H) = H
ac
(H) such that ψ = f(H)ψ (12.56)
for some f ∈ C
∞
c
((0, ∞). By the RAGE theorem the sequence ψ(t) converges
weakly to zero as t → ±∞. Abbreviate ψ(t) = exp(−itH)ψ. Introduce
ϕ
±
(t) = f(H
0
)P
±
ψ(t). (12.57)
206 12. Scattering theory
which satisfy
lim
t→±∞
ψ(t) −ϕ
+
(t) −ϕ
−
(t) = 0. (12.58)
Indeed this follows from
ψ(t) = ϕ
+
(t) +ϕ
−
(t) + (f(H) −f(H
0
))ψ(t) (12.59)
and Lemma 6.20. Moreover, we even have
lim
t→±∞
(Ω
±
−I)ϕ
±
(t) = 0 (12.60)
by Lemma 12.10. Now suppose ψ ∈ Ran(Ω
±
)
⊥
, then
ψ
2
= lim
t→±∞
'ψ(t), ψ(t)`
= lim
t→±∞
'ψ(t), ϕ
+
(t) +ϕ
−
(t)`
= lim
t→±∞
'ψ(t), Ω
+
ϕ
+
(t) + Ω
−
ϕ
−
(t)`. (12.61)
By Theorem 12.2, Ran(Ω
±
)
⊥
is invariant under H and hence ψ(t) ∈ Ran(Ω
±
)
⊥
implying
ψ
2
= lim
t→±∞
'ψ(t), Ω
∓
ϕ
∓
(t)` (12.62)
= lim
t→±∞
'P
∓
f(H
0
)
∗
Ω
∗
∓
ψ(t), ψ(t)`.
Invoking the intertwining property we see
ψ
2
= lim
t→±∞
'P
∓
f(H
0
)
∗
e
−itH
0
Ω
∗
∓
ψ, ψ(t)` = 0 (12.63)
by Corollary 12.6. Hence Ran(Ω
±
) = H
ac
(H) = H
c
(H) and we thus have
shown
Theorem 12.12 (Enß). Suppose V is short range, then the wave operators
are asymptotically complete.
For further results and references see for example [3].
Part 3
Appendix
Appendix A
Almost everything
about Lebesgue
integration
In this appendix I give a brief introduction to measure theory. Good refer
ences are [2] or [18].
A.1. Borel measures in a nut shell
The ﬁrst step in deﬁning the Lebesgue integral is extending the notion of
size from intervals to arbitrary sets. Unfortunately, this turns out to be too
much, since a classical paradox by Banach and Tarski shows that one can
break the unit ball in R
3
into a ﬁnite number of (wild – choosing the pieces
uses the Axiom of Choice and cannot be done with a jigsaw;) pieces, rotate
and translate them, and reassemble them to get two copies of the unit ball
(compare Problem A.1). Hence any reasonable notion of size (i.e., one which
is translation and rotation invariant) cannot be deﬁned for all sets!
A collection of subsets / of a given set X such that
• X ∈ /,
• / is closed under ﬁnite unions,
• / is closed under complements.
is called an algebra. Note that ∅ ∈ / and that, by de Morgan, / is also
closed under ﬁnite intersections. If an algebra is closed under countable
unions (and hence also countable intersections), it is called a σalgebra.
209
210 A. Almost everything about Lebesgue integration
Moreover, the intersection of any family of (σ)algebras ¦/
α
¦ is again
a (σ)algebra and for any collection S of subsets there is a unique smallest
(σ)algebra Σ(S) containing S (namely the intersection of all (σ)algebra
containing S). It is called the (σ)algebra generated by S.
If X is a topological space, the Borel σalgebra of X is deﬁned to be
the σalgebra generated by all open (respectively all closed) sets. Sets in the
Borel σalgebra are called Borel sets.
Example. In the case X = R
n
the Borel σalgebra will be denoted by B
n
and we will abbreviate B = B
1
.
Now let us turn to the deﬁnition of a measure: A set X together with a σ
algebra Σ is called a measure space. A measure µ is a map µ : Σ → [0, ∞]
on a σalgebra Σ such that
• µ(∅) = 0,
• µ(
¸
∞
j=1
A
j
) =
¸
∞
j=1
µ(A
j
) if A
j
∩A
k
= ∅ for all j, k (σadditivity).
It is called σﬁnite if there is a countable cover ¦X
j
¦
∞
j=1
of X with µ(X
j
) <
∞ for all j. (Note that it is no restriction to assume X
j
X.) It is called
ﬁnite if µ(X) < ∞. The sets in Σ are called measurable sets.
If we replace the σalgebra by an algebra /, then µ is called a premea
sure. In this case σadditivity clearly only needs to hold for disjoint sets
A
n
for which
¸
n
A
n
∈ /.
We will write A
n
A if A
n
⊆ A
n+1
(note A =
¸
n
A
n
) and A
n
` A if
A
n+1
⊆ A
n
(note A =
¸
n
A
n
).
Theorem A.1. Any measure µ satisﬁes the following properties:
(i) A ⊆ B implies µ(A) ≤ µ(B) (monotonicity).
(ii) µ(A
n
) → µ(A) if A
n
A (continuity from below).
(iii) µ(A
n
) → µ(A) if A
n
` A and µ(A
1
) < ∞(continuity from above).
Proof. The ﬁrst claim is obvious. The second follows using
˜
A
n
= A
n
`A
n−1
and σadditivity. The third follows from the second using
˜
A
n
= A
1
`A
n
and
µ(
˜
A
n
) = µ(A
1
) −µ(A
n
).
Example. Let A ∈ P(M) and set µ(A) to be the number of elements of A
(respectively ∞ if A is inﬁnite). This is the so called counting measure.
Note that if X = N and A
n
= ¦j ∈ N[j ≥ n¦, then µ(A
n
) = ∞, but
µ(
¸
n
A
n
) = µ(∅) = 0 which shows that the requirement µ(A
1
) < ∞ in the
last claim of Theorem A.1 is not superﬂuous.
A.1. Borel measures in a nut shell 211
A measure on the Borel σalgebra is called a Borel measure if µ(C) <
∞ for any compact set C. A Borel measures is called outer regular if
µ(A) = inf
A⊆O,O open
µ(O) (A.1)
and inner regular if
µ(A) = sup
C⊆A,C compact
µ(C). (A.2)
It is called regular if it is both outer and inner regular.
But how can we obtain some more interesting Borel measures? We will
restrict ourselves to the case of X = R for simplicity. Then the strategy
is as follows: Start with the algebra of ﬁnite unions of disjoint intervals
and deﬁne µ for those sets (as the sum over the intervals). This yields a
premeasure. Extend this to an outer measure for all subsets of R. Show
that the restriction to the Borel sets is a measure.
Let us ﬁrst show how we should deﬁne µ for intervals: To every Borel
measure on B we can assign its distribution function
µ(x) =
−µ((x, 0]), x < 0
0, x = 0
µ((0, x]), x > 0
(A.3)
which is right continuous and nondecreasing. Conversely, given a right
continuous nondecreasing function µ : R →R we can set
µ(A) =
µ(b) −µ(a), A = (a, b]
µ(b) −µ(a−), A = [a, b]
µ(b−) −µ(a), A = (a, b)
µ(b−) −µ(a−), A = [a, b)
, (A.4)
where µ(a−) = lim
ε↓0
µ(a−ε). In particular, this gives a premeasure on the
algebra of ﬁnite unions of intervals which can be extended to a measure:
Theorem A.2. For every right continuous nondecreasing function µ : R →
R there exists a unique regular Borel measure µ which extends (A.4). Two
diﬀerent functions generate the same measure if and only if they diﬀer by a
constant.
Since the proof of this theorem is rather involved, we defer it to the next
section and look at some examples ﬁrst.
Example. Suppose Θ(x) = 0 for x < 0 and Θ(x) = 1 for x ≥ 0. Then we
obtain the socalled Dirac measure at 0, which is given by Θ(A) = 1 if
0 ∈ A and Θ(A) = 0 if 0 ∈ A.
212 A. Almost everything about Lebesgue integration
Example. Suppose λ(x) = x, then the associated measure is the ordinary
Lebesgue measure on R. We will abbreviate the Lebesgue measure of a
Borel set A by λ(A) = [A[.
It can be shown that Borel measures on a separable metric space are
always regular.
A set A ∈ Σ is called a support for µ if µ(X`A) = 0. A property is
said to hold µalmost everywhere (a.e.) if the it holds on a support for
µ or, equivalently, if the set where it does not hold is contained in a set of
measure zero.
Example. The set of rational numbers has Lebesgue measure zero: λ(Q) =
0. In fact, any single point has Lebesgue measure zero, and so has any
countable union of points (by countable additivity).
Example. The Cantor set is an example of a closed uncountable set of
Lebesgue measure zero. It is constructed as follows: Start with C
0
= [0, 1]
and remove the middle third to obtain C
1
= [0,
1
3
]∪[
2
3
, 1]. Next, again remove
the middle third’s of the remaining sets to obtain C
2
= [0,
1
9
] ∪[
2
9
,
1
3
] ∪[
2
3
,
7
9
] ∪
[
8
9
, 1].
C
0
C
1
C
2
C
3
.
.
.
Proceeding like this we obtain a sequence of nesting sets C
n
and the limit
C =
¸
n
C
n
is the Cantor set. Since C
n
is compact, so is C. Moreover,
C
n
consists of 2
n
intervals of length 3
−n
, and thus its Lebesgue measure
is λ(C
n
) = (2/3)
n
. In particular, λ(C) = lim
n→∞
λ(C
n
) = 0. Using the
ternary expansion it is extremely simple to describe: C is the set of all
x ∈ [0, 1] whose ternary expansion contains no one’s, which shows that C is
uncountable (why?). It has some further interesting properties: it is totally
disconnected (i.e., it contains no subintervals) and perfect (it has no isolated
points).
Problem A.1 (Vitali set). Call two numbers x, y ∈ [0, 1) equivalent if x−y
is rational. Construct the set V by choosing one representative from each
equivalence class. Show that V cannot be measurable with respect to any
ﬁnite translation invariant measure on [0, 1). (Hint: How can you build up
[0, 1) from V ?)
A.2. Extending a premasure to a measure 213
A.2. Extending a premasure to a measure
The purpose of this section is to prove Theorem A.2. It is rather technical and
should be skipped on ﬁrst reading.
In order to prove Theorem A.2 we need to show how a premeasure can
be extended to a measure. As a prerequisite we ﬁrst establish that it suﬃces
to check increasing (or decreasing) sequences of sets when checking wether
a given algebra is in fact a σalgebra:
A collections of sets ´ is called a monotone class if A
n
A implies
A ∈ ´ whenever A
n
∈ ´ and A
n
` A implies A ∈ ´ whenever A
n
∈ ´.
Every σalgebra is a monotone class and the intersection of monotone classes
is a monotone class. Hence every collection of sets S generates a smallest
monotone class ´(S).
Theorem A.3. Let / be an algebra. Then ´(/) = Σ(/).
Proof. We ﬁrst show that ´ = ´(/) is an algebra.
Put M(A) = ¦B ∈ ´[A ∪ B ∈ ´¦. If B
n
is an increasing sequence
of sets in M(A) then A ∪ B
n
is an increasing sequence in ´ and hence
¸
n
(A∪ B
n
) ∈ ´. Now
A∪
¸
n
B
n
=
¸
n
(A∪ B
n
) (A.5)
shows that M(A) is closed under increasing sequences. Similarly, M(A) is
closed under decreasing sequences and hence it is a monotone class. But
does it contain any elements? Well if A ∈ / we have / ⊆ M(A) implying
M(A) = ´. Hence A ∪ B ∈ ´ if at least one of the sets is in /. But this
shows / ⊆ M(A) and hence M(A) = ´ for any A ∈ ´. So ´ is closed
under ﬁnite unions.
To show that we are closed under complements consider M = ¦A ∈
´[X`A ∈ ´¦. If A
n
is an increasing sequence then X`A
n
is a decreasing
sequence and X`
¸
n
A
n
=
¸
n
X`A
n
∈ ´ if A
n
∈ M. Similarly for decreas
ing sequences. Hence M is a monotone class and must be equal to ´ since
it contains /.
So we know that ´ is an algebra. To show that it is an σalgebra let
A
n
be given and put
˜
A
n
=
¸
k≤n
A
n
. Then
˜
A
n
is increasing and
¸
n
˜
A
n
=
¸
n
A
n
∈ /.
The typical use of this theorem is as follows: First verify some property
for sets in an algebra /. In order to show that it holds for any set in Σ(/),
it suﬃces to show that the sets for which it holds is closed under countable
increasing and decreasing sequences (i.e., is a monotone class).
Now we start by proving that (A.4) indeed gives rise to a premeasure.
214 A. Almost everything about Lebesgue integration
Lemma A.4. µ as deﬁned in (A.4) gives rise to a unique σﬁnite regular
premeasure on the algebra / of ﬁnite unions of disjoint intervals.
Proof. First of all, (A.4) can be extended to ﬁnite unions of disjoint inter
vals by summing over all intervals. It is straightforward to verify that µ is
well deﬁned (one set can be represented by diﬀerent unions of intervals) and
by construction additive.
To show regularity, we can assume any such union to consist of open
intervals and points only. To show outer regularity replace each point ¦x¦
by a small open interval (x+ε, x−ε) and use that µ(¦x¦) = lim
ε↓
µ(x+ε) −
µ(x−ε). Similarly, to show inner regularity, replace each open interval (a, b)
by a compact one [a
n
, b
n
] ⊆ (a, b) and use µ((a, b)) = lim
n→∞
µ(b
n
) −µ(a
n
)
if a
n
↓ a and b
n
↑ b.
It remains to verify σadditivity. We need to show
µ(
¸
k
I
k
) =
¸
k
µ(I
k
) (A.6)
whenever I
n
∈ / and I =
¸
k
I
k
∈ /. Since each I
n
is a ﬁnite union of in
tervals, we can as well assume each I
n
is just one interval (just split I
n
into
its subintervals and note that the sum does not change by additivity). Sim
ilarly, we can assume that I is just one interval (just treat each subinterval
separately).
By additivity µ is monotone and hence
n
¸
k=1
µ(I
k
) = µ(
n
¸
k=1
I
k
) ≤ µ(I) (A.7)
which shows
∞
¸
k=1
µ(I
k
) ≤ µ(I). (A.8)
To get the converse inequality we need to work harder.
By outer regularity we can cover each I
k
by open interval J
k
such that
µ(J
k
) ≤ µ(I
k
) +
ε
2
k
. Suppose I is compact ﬁrst. Then ﬁnitely many of the
J
k
, say the ﬁrst n, cover I and we have
µ(I) ≤ µ(
n
¸
k=1
J
k
) ≤
n
¸
k=1
µ(J
k
) ≤
∞
¸
k=1
µ(I
k
) +ε. (A.9)
Since ε > 0 is arbitrary, this shows σadditivity for compact intervals. By
additivity we can always add/subtract the end points of I and hence σ
additivity holds for any bounded interval. If I is unbounded, say I = [a, ∞),
A.2. Extending a premasure to a measure 215
then given x > 0 we can ﬁnd an n such that J
n
cover at least [0, x] and hence
n
¸
k=1
µ(I
k
) ≥
n
¸
k=1
µ(J
k
) −ε ≥ µ([a, x]) −ε. (A.10)
Since x > a and ε > 0 are arbitrary we are done.
This premeasure determines the corresponding measure µ uniquely (if
there is one at all):
Theorem A.5 (Uniqueness of measures). Let µ be a σﬁnite premeasure
on an algebra /. Then there is at most one extension to Σ(/).
Proof. We ﬁrst assume that µ(X) < ∞. Suppose there is another extension
˜ µ and consider the set
S = ¦A ∈ Σ(/)[µ(A) = ˜ µ(A)¦. (A.11)
I claim S is a monotone class and hence S = Σ(/) since / ⊆ S by assump
tion (Theorem A.3).
Let A
n
A. If A
n
∈ S we have µ(A
n
) = ˜ µ(A
n
) and taking limits
(Theorem A.1 (ii)) we conclude µ(A) = ˜ µ(A). Next let A
n
` A and take
again limits. This ﬁnishes the ﬁnite case. To extend our result to the σﬁnite
case let X
j
X be an increasing sequence such that µ(X
j
) < ∞. By the
ﬁnite case µ(A∩ X
j
) = ˜ µ(A∩ X
j
) (just restrict µ, ˜ µ to X
j
). Hence
µ(A) = lim
j→∞
µ(A∩ X
j
) = lim
j→∞
˜ µ(A∩ X
j
) = ˜ µ(A) (A.12)
and we are done.
Note that if our premeasure is regular, so will be the extension:
Lemma A.6. Suppose µ is a σﬁnite premeasure on some algebra / gen
erating the Borel sets B. Then outer (inner) regularity holds for all Borel
sets if it holds for all sets in /.
Proof. We ﬁrst assume that µ(X) < ∞. Set
µ
◦
(A) = inf
A⊆O,O open
µ(O) ≥ µ(A) (A.13)
and let M = ¦A ∈ B[µ
◦
(A) = µ(A)¦. Since by assumption M contains
some algebra generating B it suﬃces to proof that M is a monotone class.
Let A
n
∈ M be a monotone sequence and let O
n
⊇ A
n
be open sets such
that µ(O
n
) ≤ µ(A
n
) +
1
n
. Then
µ(A
n
) ≤ µ
◦
(A
n
) ≤ µ(O
n
) ≤ µ(A
n
) +
1
n
. (A.14)
Now if A
n
A just take limits and use continuity from below of µ. Similarly
if A
n
` A.
216 A. Almost everything about Lebesgue integration
Now let µ be arbitrary. Given A we can split it into disjoint sets A
j
such that A
j
⊆ X
j
(A
1
= A ∩ X
1
, A
2
= (A`A
1
) ∩ X
2
, etc.). Let X
j
be a
cover with µ(X
j
) < ∞. By regularity, we can assume X
j
open. Thus there
are open (in X) sets O
j
covering A
j
such that µ(O
j
) ≤ µ(A
j
) +
ε
2
j
. Then
O =
¸
j
O
j
is open, covers A, and satisﬁes
µ(A) ≤ µ(O) ≤
¸
j
µ(O
j
) ≤ µ(A) +ε. (A.15)
This settles outer regularity.
Next let us turn to inner regularity. If µ(X) < ∞one can show as before
that M = ¦A ∈ B[µ
◦
(A) = µ(A)¦, where
µ
◦
(A) = sup
C⊆A,C compact
µ(C) ≤ µ(A) (A.16)
is a monotone class. This settles the ﬁnite case.
For the σﬁnite case split again A as before. Since X
j
has ﬁnite measure,
there are compact subsets K
j
of A
j
such that µ(A
j
) ≤ µ(K
j
) +
ε
2
j
. Now
we need to distinguish two cases: If µ(A) = ∞, the sum
¸
j
µ(A
j
) will
diverge and so will
¸
j
µ(K
j
). Hence
˜
K
n
=
¸
n
j=1
⊆ A is compact with
µ(
˜
K
n
) → ∞ = µ(A). If µ(A) < ∞, the sum
¸
j
µ(A
j
) will converge and
choosing n suﬃciently large we will have
µ(
˜
K
n
) ≤ µ(A) ≤ µ(
˜
K
n
) + 2ε. (A.17)
This ﬁnishes the proof.
So it remains to ensure that there is an extension at all. For any pre
measure µ we deﬁne
µ
∗
(A) = inf
∞
¸
n=1
µ(A
n
)
A ⊆
∞
¸
n=1
A
n
, A
n
∈ /
¸
(A.18)
where the inﬁmum extends over all countable covers from /. Then the
function µ
∗
: P(X) → [0, ∞] is an outer measure, that is, it has the
properties (Problem A.2)
• µ
∗
(∅) = 0,
• A
1
⊆ A
2
⇒ µ
∗
(A
1
) ≤ µ
∗
(A
2
), and
• µ
∗
(
¸
∞
n=1
A
n
) ≤
¸
∞
n=1
µ
∗
(A
n
) (subadditivity).
Note that µ
∗
(A) = µ(A) for A ∈ / (Problem A.3).
Theorem A.7 (Extensions via outer measures). Let µ
∗
be an outer measure.
Then the set Σ of all sets A satisfying the Carath´eodory condition
µ
∗
(E) = µ
∗
(A∩ E) +µ
∗
(A
∩ E) ∀E ⊆ X (A.19)
A.2. Extending a premasure to a measure 217
(where A
= X`A is the complement of A) form a σalgebra and µ
∗
restricted
to this σalgebra is a measure.
Proof. We ﬁrst show that Σ is an algebra. It clearly contains X and is closed
under complements. Let A, B ∈ Σ. Applying Carath´eodory’s condition
twice ﬁnally shows
µ
∗
(E) = µ
∗
(A∩ B ∩ E) +µ
∗
(A
∩ B ∩ E) +µ
∗
(A∩ B
∩ E)
+µ
∗
(A
∩ B
∩ E)
≥ µ
∗
((A∪ B) ∩ E) +µ
∗
((A∪ B)
∩ E), (A.20)
where we have used De Morgan and
µ
∗
(A∩B∩E) +µ
∗
(A
∩B∩E) +µ
∗
(A∩B
∩E) ≥ µ
∗
((A∪B) ∩E) (A.21)
which follows from subadditivity and (A ∪ B) ∩ E = (A ∩ B ∩ E) ∪ (A
∩
B ∩ E) ∪ (A∩ B
∩ E). Since the reverse inequality is just subadditivity, we
conclude that Σ is an algebra.
Next, let A
n
be a sequence of sets from Σ. Without restriction we
can assume that they are disjoint (compare the last argument in proof of
Theorem A.3). Abbreviate
˜
A
n
=
¸
k≤n
A
n
, A =
¸
n
A
n
. Then for any set E
we have
µ
∗
(
˜
A
n
∩ E) = µ
∗
(A
n
∩
˜
A
n
∩ E) +µ
∗
(A
n
∩
˜
A
n
∩ E)
= µ
∗
(A
n
∩ E) +µ
∗
(
˜
A
n−1
∩ E)
= . . . =
n
¸
k=1
µ
∗
(A
k
∩ E). (A.22)
Using
˜
A
n
∈ Σ and monotonicity of µ
∗
, we infer
µ
∗
(E) = µ
∗
(
˜
A
n
∩ E) +µ
∗
(
˜
A
n
∩ E)
≥
n
¸
k=1
µ
∗
(A
k
∩ E) +µ
∗
(A
∩ E). (A.23)
Letting n → ∞ and using subadditivity ﬁnally gives
µ
∗
(E) ≥
∞
¸
k=1
µ
∗
(A
k
∩ E) +µ
∗
(A
∩ E)
≥ µ
∗
(A∩ E) +µ
∗
(B
∩ E) ≥ µ
∗
(E) (A.24)
and we infer that Σ is a σalgebra.
Finally, setting E = A in (A.24) we have
µ
∗
(A) =
∞
¸
k=1
µ
∗
(A
k
∩ A) +µ
∗
(A
∩ A) =
∞
¸
k=1
µ
∗
(A
k
) (A.25)
and we are done.
218 A. Almost everything about Lebesgue integration
Remark: The constructed measure µ is complete, that is, for any mea
surable set A of measure zero, any subset of A is again measurable (Prob
lem A.4).
The only remaining question is wether there are any nontrivial sets sat
isfying the Carath´eodory condition.
Lemma A.8. Let µ be a premeasure on / and let µ
∗
be the associated outer
measure. Then every set in / satisﬁes the Carath´eodory condition.
Proof. Let A
n
∈ / be a countable cover for E. Then for any A ∈ / we
have
∞
¸
n=1
µ(A
n
) =
∞
¸
n=1
µ(A
n
∩A)+
∞
¸
n=1
µ(A
n
∩A
) ≥ µ
∗
(E∩A)+µ
∗
(E∩A
) (A.26)
since A
n
∩A ∈ / is a cover for E ∩A and A
n
∩A
∈ / is a cover for E ∩A
.
Taking the inﬁmum we have µ
∗
(E) ≥ µ
∗
(E∩A) +µ
∗
(E∩A
) which ﬁnishes
the proof.
Thus, as a consequence we obtain Theorem A.2.
Problem A.2. Show that µ
∗
deﬁned in (A.18) is an outer measure. (Hint
for the last property: Take a cover ¦B
nk
¦
∞
k=1
for A
n
such that µ
∗
(A
n
) =
ε
2
n
+
¸
∞
k=1
µ(B
nk
) and note that ¦B
nk
¦
∞
n,k=1
is a cover for
¸
n
A
n
.)
Problem A.3. Show that µ
∗
deﬁned in (A.18) extends µ. (Hint: For the
cover A
n
it is no restriction to assume A
n
∩ A
m
= ∅ and A
n
⊆ A.)
Problem A.4. Show that the measure constructed in Theorem A.7 is com
plete.
A.3. Measurable functions
The Riemann integral works by splitting the x coordinate into small intervals
and approximating f(x) on each interval by its minimum and maximum.
The problem with this approach is that the diﬀerence between maximum
and minimum will only tend to zero (as the intervals get smaller) if f(x) is
suﬃciently nice. To avoid this problem we can force the diﬀerence to go to
zero by considering, instead of an interval, the set of x for which f(x) lies
between two given numbers a < b. Now we need the size of the set of these
x, that is, the size of the preimage f
−1
((a, b)). For this to work, preimages
of intervals must be measurable.
A function f : X → R
n
is called measurable if f
−1
(A) ∈ Σ for every
A ∈ B
n
. A complexvalued function is called measurable if both its real
and imaginary parts are. Clearly it suﬃces to check this condition for every
A.3. Measurable functions 219
set A in a collection of sets which generate B
n
, say for all open intervals or
even for open intervals which are products of (a, ∞):
Lemma A.9. A function f : X →R
n
is measurable if and only if
f
−1
(I) ∈ Σ ∀ I =
n
¸
j=1
(a
j
, ∞). (A.27)
In particular, a function f : X → R
n
is measurable if and only if every
component is measurable.
Proof. All you have to use is f
−1
(R
n
`A) = X`f
−1
(A), f
−1
(
¸
j
A
j
) =
¸
j
f
1
(A
j
) and the fact that any open set is a countable union of open
intervals.
If Σ is the Borel σalgebra, we will call a measurable function also Borel
function. Note that, in particular,
Lemma A.10. Any continuous function is measurable and the composition
of two measurable functions is again measurable.
Moreover, sometimes it is also convenient to allow ±∞as possible values
for f, that is, functions f : X → R, R = R ∪ ¦−∞, ∞¦. In this case A ⊆ R
is called Borel if A∩ R is.
The set of all measurable functions forms an algebra.
Lemma A.11. Suppose f, g : X → R are measurable functions. Then the
sum f +g and the product fg is measurable.
Proof. Note that addition and multiplication are continuous functions from
R
2
→R and hence the claim follows from the previous lemma.
Moreover, the set of all measurable functions is closed under all impor
tant limiting operations.
Lemma A.12. Suppose f
n
: X →R is a sequence of measurable functions,
then
inf
n∈N
f
n
, sup
n∈N
f
n
, liminf
n→∞
f
n
, limsup
n→∞
f
n
(A.28)
are measurable as well.
Proof. It suﬃces to proof that sup f
n
is measurable since the rest follows
from inf f
n
= −sup(−f
n
), liminf f
n
= sup
k
inf
n≥k
f
n
, and limsup f
n
=
inf
k
sup
n≥k
f
n
. But (sup f
n
)
−1
((a, ∞)) =
¸
n
f
−1
n
((a, ∞)) and we are done.
It follows that if f and g are measurable functions, so are min(f, g),
max(f, g), [f[ = max(f, −f), f
±
= max(±f, 0).
220 A. Almost everything about Lebesgue integration
A.4. The Lebesgue integral
Now we can deﬁne the integral for measurable functions as follows. A
measurable function s : X → R is called simple if its range is ﬁnite
s(X) = ¦α
j
¦
p
j=1
, that is, if
s =
p
¸
j=1
α
j
χ
A
j
, A
j
= s
−1
(α
j
) ∈ Σ. (A.29)
Here χ
A
is the characteristic function of A, that is, χ
A
(x) = 1 if x ∈ A
and χ
A
(x) = 0 else.
For a positive simple function we deﬁne its integral as
A
s dµ =
n
¸
j=1
α
j
µ(A
j
∩ A). (A.30)
Here we use the convention 0 ∞ = 0.
Lemma A.13. The integral has the following properties:
(i)
A
s dµ =
X
χ
A
s dµ.
(ii)
S
∞
j=1
A
j
s dµ =
¸
∞
j=1
A
j
s dµ.
(iii)
A
αs dµ = α
A
s dµ.
(iv)
A
(s +t)dµ =
A
s dµ +
A
t dµ.
(v) A ⊆ B ⇒
A
s dµ ≤
B
s dµ.
(vi) s ≤ t ⇒
A
s dµ ≤
A
t dµ.
Proof. (i) is clear from the deﬁnition. (ii) follows from σadditivity of µ.
(iii) is obvious. (iv) Let s =
¸
j
α
j
χ
A
j
, t =
¸
j
β
j
χ
B
j
and abbreviate
C
jk
= (A
j
∩ B
k
) ∩ A. Then
A
(s +t)dµ =
¸
j,k
C
jk
(s +t)dµ =
¸
j,k
(α
j
+β
k
)µ(C
jk
)
=
¸
j,k
C
jk
s dµ +
C
jk
t dµ
=
A
s dµ +
A
t dµ(A.31)
(v) follows from monotonicity of µ. (vi) follows using t −s ≥ 0 and arguing
as in (iii).
Our next task is to extend this deﬁnition to arbitrary positive functions
by
A
f dµ = sup
s≤f
A
s dµ, (A.32)
A.4. The Lebesgue integral 221
where the supremum is taken over all simple functions s ≤ f. Note that,
except for possibly (ii) and (iv), Lemma A.13 still holds for this extension.
Theorem A.14 (monotone convergence). Let f
n
be a monotone nondecreas
ing sequence of positive measurable functions, f
n
f. Then
A
f
n
dµ →
A
f dµ. (A.33)
Proof. By property (v)
A
f
n
dµ is monotone and converges to some number
α. By f
n
≤ f and again (v) we have
α ≤
A
f dµ. (A.34)
To show the converse let s be simple such that s ≤ f and let θ ∈ (0, 1). Put
A
n
= ¦x ∈ A[f
n
(x) ≥ θs(x)¦ and note A
n
X (show this). Then
A
f
n
dµ ≥
An
f
n
dµ ≥ θ
An
s dµ. (A.35)
Letting n → ∞ we see
α ≥ θ
A
s dµ. (A.36)
Since this is valid for any θ < 1, it still holds for θ = 1. Finally, since s ≤ f
is arbitrary, the claim follows.
In particular
A
f dµ = lim
n→∞
A
s
n
dµ, (A.37)
for any monotone sequence s
n
f of simple functions. Note that there is
always such a sequence, for example,
s
n
(x) =
n
2
¸
k=0
k
n
χ
f
−1
(A
k
)
(x), A
k
= [
k
n
,
k + 1
n
2
), A
n
2 = [n, ∞). (A.38)
By construction s
n
converges uniformly if f is bounded, since s
n
(x) = n if
f(x) = ∞ and f(x) −s
n
(x) <
1
n
if f(x) < n + 1.
Now what about the missing items (ii) and (iv) from Lemma A.13? Since
limits can be spread over sums, the extension is linear (i.e., item (iv) holds)
and (ii) also follows directly from the monotone convergence theorem. We
even have the following result:
Lemma A.15. If f ≥ 0 is measurable, then dν = f dµ deﬁned via
ν(A) =
A
f dµ (A.39)
222 A. Almost everything about Lebesgue integration
is a measure such that
g dν =
gf dµ. (A.40)
Proof. As already mentioned, additivity of µ is equivalent to linearity of the
integral and σadditivity follows from the monotone convergence theorem
ν(
∞
¸
n=1
A
n
) =
(
∞
¸
n=1
χ
An
)f dµ =
∞
¸
n=1
χ
An
f dµ =
∞
¸
n=1
ν(A
n
). (A.41)
The second claim holds for simple functions and hence for all functions by
construction of the integral.
If f
n
is not necessarily monotone we have at least
Theorem A.16 (Fatou’s Lemma). If f
n
is a sequence of nonnegative mea
surable function, then
A
liminf
n→∞
f
n
dµ ≤ liminf
n→∞
A
f
n
dµ, (A.42)
Proof. Set g
n
= inf
k≥n
f
k
. Then g
n
≤ f
n
implying
A
g
n
dµ ≤
A
f
n
dµ. (A.43)
Now take the liminf on both sides and note that by the monotone conver
gence theorem
liminf
n→∞
A
g
n
dµ = lim
n→∞
A
g
n
dµ =
A
lim
n→∞
g
n
dµ =
A
liminf
n→∞
f
n
dµ,
(A.44)
proving the claim.
If the integral is ﬁnite for both the positive and negative part f
±
of an
arbitrary measurable function f, we call f integrable and set
A
f dµ =
A
f
+
dµ −
A
f
−
dµ. (A.45)
The set of all integrable functions is denoted by L
1
(X, dµ).
Lemma A.17. Lemma A.13 holds for integrable functions s, t.
Similarly, we handle the case where f is complexvalued by calling f
integrable if both the real and imaginary part are and setting
A
f dµ =
A
Re(f)dµ + i
A
Im(f)dµ. (A.46)
Clearly f is integrable if and only if [f[ is.
A.4. The Lebesgue integral 223
Lemma A.18. For any integrable functions f, g we have
[
A
f dµ[ ≤
A
[f[ dµ (A.47)
and (triangle inequality)
A
[f +g[ dµ ≤
A
[f[ dµ +
A
[g[ dµ. (A.48)
Proof. Put α =
z
∗
z
, where z =
A
f dµ (without restriction z = 0). Then
[
A
f dµ[ = α
A
f dµ =
A
αf dµ =
A
Re(αf) dµ ≤
A
[f[ dµ. (A.49)
proving the ﬁrst claim. The second follows from [f +g[ ≤ [f[ +[g[.
In addition, our integral is well behaved with respect to limiting opera
tions.
Theorem A.19 (dominated convergence). Let f
n
be a convergent sequence
of measurable functions and set f = lim
n→∞
f
n
. Suppose there is an inte
grable function g such that [f
n
[ ≤ g. Then f is integrable and
lim
n→∞
f
n
dµ =
fdµ. (A.50)
Proof. The real and imaginary parts satisfy the same assumptions and so
do the positive and negative parts. Hence it suﬃces to prove the case where
f
n
and f are nonnegative.
By Fatou’s lemma
liminf
n→∞
A
f
n
dµ ≥
A
f dµ (A.51)
and
liminf
n→∞
A
(g −f
n
)dµ ≥
A
(g −f)dµ. (A.52)
Subtracting
A
g dµ on both sides of the last inequality ﬁnishes the proof
since liminf(−f
n
) = −limsup f
n
.
Remark: Since sets of measure zero do not contribute to the value of the
integral, it clearly suﬃces if the requirements of the dominated convergence
theorem are satisﬁed almost everywhere (with respect to µ).
Note that the existence of g is crucial, as the example f
n
(x) =
1
n
χ
[−n,n]
(x)
on R with Lebesgue measure shows.
224 A. Almost everything about Lebesgue integration
Example. If µ(x) =
¸
n
α
n
Θ(x − x
n
) is a sum of Dirac measures Θ(x)
centered at x = 0, then
f(x)dµ(x) =
¸
n
α
n
f(x
n
). (A.53)
Hence our integral contains sums as special cases.
Problem A.5. Show that the set B(X) of bounded measurable functions
is a Banach space. Show that the set S(X) of simple functions is dense
in B(X). Show that the integral is a bounded linear functional on B(X).
(Hence Theorem 0.24 could be used to extend the integral from simple to
bounded measurable functions.)
Problem A.6. Show that the dominated convergence theorem implies (un
der the same assumptions)
lim
n→∞
[f
n
−f[dµ = 0. (A.54)
Problem A.7. Suppose y → f(x, y) is measurable for every x and x →
f(x, y) is continuous for every y. Show that
F(x) =
A
f(x, y) dµ(y) (A.55)
is continuous if there is an integrable function g(y) such that [f(x, y)[ ≤ g(y).
Problem A.8. Suppose y → f(x, y) is measurable for ﬁxed x and x →
f(x, y) is diﬀerentiable for ﬁxed y. Show that
F(x) =
A
f(x, y) dµ(y) (A.56)
is diﬀerentiable if there is an integrable function g(y) such that [
∂
∂x
f(x, y)[ ≤
g(y). Moreover, x →
∂
∂x
f(x, y) is measurable and
F
(x) =
A
∂
∂x
f(x, y) dµ(y) (A.57)
in this case.
A.5. Product measures
Let µ
1
and µ
2
be two measures on Σ
1
and Σ
2
, respectively. Let Σ
1
⊗Σ
2
be
the σalgebra generated by rectangles of the form A
1
A
2
.
Example. Let B be the Borel sets in R then B
2
= B⊗ B are the Borel
sets in R
2
(since the rectangles are a basis for the product topology).
Any set in Σ
1
⊗Σ
2
has the section property, that is,
A.5. Product measures 225
Lemma A.20. Suppose A ∈ Σ
1
⊗Σ
2
then its sections
A
1
(x
2
) = ¦x
1
[(x
1
, x
2
) ∈ A¦ and A
2
(x
1
) = ¦x
2
[(x
1
, x
2
) ∈ A¦ (A.58)
are measurable.
Proof. Denote all sets A ∈ Σ
1
⊗Σ
2
in with the property that A
1
(x
2
) ∈ Σ
1
by
S. Clearly all rectangles are in S and it suﬃces to show that S is a σalgebra.
Moreover, if A ∈ S, then (A
)
1
(x
2
) = (A
1
(x
2
))
∈ Σ
2
and thus S is closed
under complements. Similarly, if A
n
∈ S, then (
¸
n
A
n
)
1
(x
2
) =
¸
n
(A
n
)
1
(x
2
)
shows that S is closed under countable unions.
This implies that if f is a measurable function on X
1
X
2
, then f(., x
2
) is
measurable on X
1
for every x
2
and f(x
1
, .) is measurable on X
2
for every x
1
(observe A
1
(x
2
) = ¦x
1
[f(x
1
, x
2
) ∈ B¦, where A = ¦(x
1
, x
2
)[f(x
1
, x
2
) ∈ B¦).
In fact, this is even equivalent since χ
A
1
(x
2
)
(x
1
) = χ
A
2
(x
1
)
(x
2
) = χ
A
(x
1
, x
2
).
Given two measures µ
1
on Σ
1
and µ
2
on Σ
2
we now want to construct
the product measure, µ
1
⊗µ
2
on Σ
1
⊗Σ
2
such that
µ
1
⊗µ
2
(A
1
A
2
) = µ
1
(A
1
)µ
2
(A
2
), A
j
∈ Σ
j
, j = 1, 2. (A.59)
Theorem A.21. Let µ
1
and µ
2
be two σﬁnite measures on Σ
1
and Σ
2
,
respectively. Let A ∈ Σ
1
⊗ Σ
2
. Then µ
2
(A
2
(x
1
)) and µ
1
(A
1
(x
2
)) are mea
surable and
X
1
µ
2
(A
2
(x
1
))dµ
1
(x
1
) =
X
2
µ
1
(A
1
(x
2
))dµ
2
(x
2
). (A.60)
Proof. Let S be the set of all subsets for which our claim holds. Note
that S contains at least all rectangles. It even contains the algebra of ﬁnite
disjoint unions of rectangles. Thus it suﬃces to show that S is a monotone
class. If µ
1
and µ
2
are ﬁnite, this follows from continuity from above and
below of measures. The case if µ
1
and µ
2
are σﬁnite can be handled as in
Theorem A.5.
Hence we can deﬁne
µ
1
⊗µ
2
(A) =
X
1
µ
2
(A
2
(x
1
))dµ
1
(x
1
) =
X
2
µ
1
(A
1
(x
2
))dµ
2
(x
2
) (A.61)
or equivalently
µ
1
⊗µ
2
(A) =
X
1
X
2
χ
A
(x
1
, x
2
)dµ
2
(x
2
)
dµ
1
(x
1
)
=
X
2
X
1
χ
A
(x
1
, x
2
)dµ
1
(x
1
)
dµ
2
(x
2
). (A.62)
Additivity of µ
1
⊗µ
2
follows from the monotone convergence theorem.
226 A. Almost everything about Lebesgue integration
Note that (A.59) uniquely deﬁnes µ
1
⊗ µ
2
as a σﬁnite premeasure on
the algebra of ﬁnite disjoint unions of rectangles. Hence by Theorem A.5 it
is the only measure on Σ
1
⊗Σ
2
satisfying (A.59).
Finally we have:
Theorem A.22 (Fubini). Let f be a measurable function on X
1
X
2
and
let µ
1
, µ
2
be σﬁnite measures on X
1
, X
2
, respectively.
(i) If f ≥ 0 then
f(., x
2
)dµ
2
(x
2
) and
f(x
1
, .)dµ
1
(x
1
) are both mea
surable and
f(x
1
, x
2
)dµ
1
⊗µ
2
(x
1
, x
2
) =
f(x
1
, x
2
)dµ
1
(x
1
)
dµ
2
(x
2
)
=
f(x
1
, x
2
)dµ
2
(x
2
)
dµ
1
(x
1
). (A.63)
(ii) If f is complex then
[f(x
1
, x
2
)[dµ
1
(x
1
) ∈ L
1
(X
2
, dµ
2
) (A.64)
respectively
[f(x
1
, x
2
)[dµ
2
(x
2
) ∈ L
1
(X
1
, dµ
1
) (A.65)
if and only if f ∈ L
1
(X
1
X
2
, dµ
1
⊗ dµ
2
). In this case (A.63)
holds.
Proof. By Theorem A.21 the claim holds for simple functions. Now (i)
follows from the monotone convergence theorem and (ii) from the dominated
convergence theorem.
In particular, if f(x
1
, x
2
) is either nonnegative or integrable, then the
order of integration can be interchanged.
Lemma A.23. If µ
1
and µ
2
are σﬁnite regular Borel measures with, so is
µ
1
⊗µ
2
.
Proof. Regularity holds for every rectangle and hence also for the algebra of
ﬁnite disjoint unions of rectangles. Thus the claim follows from Lemma A.6.
Note that we can iterate this procedure.
Lemma A.24. Suppose µ
j
, j = 1, 2, 3 are σﬁnite measures. Then
(µ
1
⊗µ
2
) ⊗µ
3
= µ
1
⊗(µ
2
⊗µ
3
). (A.66)
A.6. Decomposition of measures 227
Proof. First of all note that (Σ
1
⊗Σ
2
) ⊗Σ
3
= Σ
1
⊗(Σ
2
⊗Σ
3
) is the sigma
algebra generated by the cuboids A
1
A
2
A
3
in X
1
X
2
X
3
. Moreover,
since
((µ
1
⊗µ
2
) ⊗µ
3
)(A
1
A
2
A
3
) = µ
1
(A
1
)µ
2
(A
2
)µ
3
(A
3
)
= (µ
1
⊗(µ
2
⊗µ
3
))(A
1
A
2
A
3
) (A.67)
the two measures coincide on the algebra of ﬁnite disjoint unions of cuboids.
Hence they coincide everywhere by Theorem A.5.
Example. If λ is Lebesgue measure on R, then λ
n
= λ⊗ ⊗λ is Lebesgue
measure on R
n
. Since λ is regular, so is λ
n
.
A.6. Decomposition of measures
Let µ, ν be two measures on a measure space (X, Σ). They are called
mutually singular (in symbols µ ⊥ ν) if they are supported on disjoint
sets. That is, there is a measurable set N such that µ(N) = 0 and ν(X`N) =
0.
Example. Let λ be the Lebesgue measure and Θ the Dirac measure
(centered at 0), then λ ⊥ Θ: Just take N = ¦0¦, then λ(¦0¦) = 0 and
Θ(R`¦0¦) = 0.
On the other hand, ν is called absolutely continuous with respect to
µ (in symbols ν < µ) if µ(A) = 0 implies ν(A) = 0.
Example. The prototypical example is the measure dν = f dµ (compare
Lemma A.15). Indeed µ(A) = 0 implies
ν(A) =
A
f dµ = 0 (A.68)
and shows that ν is absolutely continuous with respect to µ. In fact, we will
show below that every absolutely continuous measure is of this form.
The two main results will follow as simple consequence of the following
result:
Theorem A.25. Let µ, ν be σﬁnite measures. Then there exists a unique
(a.e.) nonnegative function f and a set N of µ measure zero, such that
ν(A) = ν(A∩ N) +
A
f dµ. (A.69)
Proof. We ﬁrst assume µ, ν to be ﬁnite measures. Let α = µ + ν and
consider the Hilbert space L
2
(X, dα). Then
(h) =
X
hdν (A.70)
228 A. Almost everything about Lebesgue integration
is a bounded linear functional by CauchySchwarz:
[(h)[
2
=
X
1 hdν
2
≤
[1[
2
dν
[h[
2
dν
≤ ν(X)
[h[
2
dα
= ν(X)h
2
. (A.71)
Hence by the Riesz lemma (Theorem 1.7) there exists an g ∈ L
2
(X, dα) such
that
(h) =
X
hg dα. (A.72)
By construction
ν(A) =
χ
A
dν =
χ
A
g dα =
A
g dα. (A.73)
In particular, g must be positive a.e. (take A the set where g is negative).
Furthermore, let N = ¦x[g(x) ≥ 1¦, then
ν(N) =
N
g dα ≥ α(N) = µ(N) +ν(N), (A.74)
which shows µ(N) = 0. Now set
f =
g
1 −g
χ
N
, N
= X`N. (A.75)
Then, since (A.73) implies dν = g dα respectively dµ = (1 −g)dα, we have
A
fdµ =
χ
A
g
1 −g
χ
N
dµ
=
χ
A∩N
g dα
= ν(A∩ N
) (A.76)
as desired. Clearly f is unique, since if there is a second function
˜
f, then
A
(f −
˜
f)dµ = 0 for every A shows f −
˜
f = 0 a.e..
To see the σﬁnite case, observe that X
n
X, µ(X
n
) < ∞and Y
n
X,
ν(Y
n
) < ∞ implies X
n
∩ Y
n
X and α(X
n
∩ Y
n
) < ∞. Hence when
restricted to X
n
∩Y
n
we have sets N
n
and functions f
n
. Now take N =
¸
N
n
and choose f such that f[
Xn
= f
n
(this is possible since f
n+1
[
Xn
= f
n
a.e.).
Then µ(N) = 0 and
ν(A∩ N
) = lim
n→∞
ν(A∩ (X
n
`N)) = lim
n→∞
A∩Xn
f dµ =
A
f dµ, (A.77)
which ﬁnishes the proof.
Now the anticipated results follow with no eﬀort:
A.7. Derivatives of measures 229
Theorem A.26 (Lebesgue decomposition). Let µ, ν be two σﬁnite mea
sures on a measure space (X, Σ). Then ν can be uniquely decomposed as
ν = ν
ac
+ ν
sing
, where ν
ac
and ν
sing
are mutually singular and ν
ac
is abso
lutely continuous with respect to µ.
Proof. Taking ν
sing
(A) = ν(A ∩ N) and dν
ac
= f dµ there is at least one
such decomposition. To show uniqueness, let ν be ﬁnite ﬁrst. If there
is another one ν = ˜ ν
ac
+ ˜ ν
sing
, then let
˜
N be such that µ(
˜
N) = 0 and
˜ ν
sing
(
˜
N
) = 0. Then ˜ ν
sing
(A) − ˜ ν
sing
(A) =
A
(
˜
f − f)dµ. In particular,
A∩N
∩
˜
N
(
˜
f − f)dµ = 0 and hence
˜
f = f a.e. away from N ∪
˜
N. Since
µ(N ∪
˜
N) = 0, we have
˜
f = f a.e. and hence ˜ ν
ac
= ν
ac
as well as ˜ ν
sing
=
ν − ˜ ν
ac
= ν −ν
ac
= ν
sing
. The σﬁnite case follows as usual.
Theorem A.27 (RadonNikodym). Let µ, ν be two σﬁnite measures on a
measure space (X, Σ). Then ν is absolutely continuous with respect to µ if
and only if there is a positive measurable function f such that
ν(A) =
A
f dµ (A.78)
for every A ∈ Σ. The function f is determined uniquely a.e. with respect to
µ and is called the RadonNikodym derivative
dν
dµ
of ν with respect to µ.
Proof. Just observe that in this case ν(A ∩ N) = 0 for every A, that is
ν
sing
= 0.
Problem A.9. Let µ is a Borel measure on B and suppose its distribution
function µ(x) is diﬀerentiable. Show that the RadonNikodym derivative
equals the ordinary derivative µ
(x).
Problem A.10 (Chain rule). Show that ν < µ is a transitive relation. In
particular, if ω < ν < µ show that
dω
dµ
=
dω
dν
dν
dµ
.
Problem A.11. Suppose ν < µ. Show that
dω
dµ
dµ =
dω
dν
dν +dζ,
where ζ is a positive measure which is singular with respect to ν. Show that
ζ = 0 if and only if µ < ν.
A.7. Derivatives of measures
If µ is a Borel measure on B and its distribution function µ(x) is diﬀeren
tiable, then the RadonNikodym derivative is just the ordinary derivative
230 A. Almost everything about Lebesgue integration
µ
(x). Our aim in this section is to generalize this result to arbitrary mea
sures on B
n
.
We call
(Dµ)(x) = lim
ε↓0
µ(B
ε
(x))
[B
ε
(x)[
, (A.79)
the derivative of µ at x ∈ R
n
provided the above limit exists. (Here B
r
(x) ⊂
R
n
is a ball of radius r centered at x ∈ R
n
and [A[ denotes the Lebesgue
measure of A ∈ B
n
).
Note that for a Borel measure on B, (Dµ)(x) exists if and only if µ(x)
(as deﬁned in (A.3)) is diﬀerentiable at x and (Dµ)(x) = µ
(x) in this case.
We will assume that µ is regular throughout this section. It can be
shown that every Borel
To compute the derivative of µ we introduce the upper and lower
derivative
(Dµ)(x) = limsup
ε↓0
µ(B
ε
(x))
[B
ε
(x)[
and (Dµ)(x) = liminf
ε↓0
µ(B
ε
(x))
[B
ε
(x)[
. (A.80)
Clearly µ is diﬀerentiable if (Dµ)(x) = (Dµ)(x) < ∞. First of all note that
they are measurable:
Lemma A.28. The upper derivative is lower semicontinuous, that is the set
¦x[(Dµ)(x) > α¦ is open for every α ∈ R. Similarly, the lower derivative is
upper semicontinuous, that is ¦x[(Dµ)(x) < α¦ is open.
Proof. We only prove the claim for Dµ, the case Dµ being similar. Abbre
viate,
M
r
(x) = sup
0<ε<r
µ(B
ε
(x))
[B
ε
(x)[
(A.81)
and note that it suﬃces to show that O
r
= ¦x[M
r
(x) > α¦ is open.
If x ∈ O
r
, there is some ε < r such that
µ(B
ε
(x))
[B
ε
(x)[
> α. (A.82)
Let δ > 0 and y ∈ B
δ
(x). Then B
ε
(x) ⊆ B
ε+δ
(y) implying
µ(B
ε+δ
(y))
[B
ε+δ
(y)[
≥
ε
ε +δ
n
µ(B
ε
(x))
[B
ε
(x)[
> α (A.83)
for δ suﬃciently small. That is, B
δ
(x) ⊆ O.
In particular, both the upper and lower derivative are measurable. Next,
the following geometric fact of R
n
will be needed.
A.7. Derivatives of measures 231
Lemma A.29. Given open balls B
1
, . . . , B
m
in R
n
, there is a subset of
disjoint balls B
j
1
, . . . , B
j
k
such that
m
¸
i=1
B
i
≤ 3
n
k
¸
i=1
[B
j
i
[. (A.84)
Proof. Start with B
j
1
= B
1
= B
r
1
(x
1
) and remove all balls from our list
which intersect B
j
1
. Observe that the removed balls are all contained in
3B
1
= B
3r
1
(x
1
). Proceeding like this we obtain B
j
1
, . . . , B
j
k
such that
m
¸
i=1
B
i
⊆
k
¸
i=1
B
3r
j
i
(x
j
i
) (A.85)
and the claim follows since [B
3r
(x)[ = 3
n
[B
r
(x)[.
Now we can show
Lemma A.30. Let α > 0. For any Borel set A we have
[¦x ∈ A[ (Dµ)(x) > α¦[ ≤ 3
n
µ(A)
α
(A.86)
and
[¦x ∈ A[ (Dµ)(x) > 0¦[ = 0, whenever µ(A) = 0. (A.87)
Proof. Let A
α
= ¦x[(Dµ)(x) > α¦. We will show
[K[ ≤ 3
n
µ(O)
α
(A.88)
for any compact set K and open set O with K ⊆ E ⊆ O. The ﬁrst claim
then follows from regularity of µ and the Lebesgue measure.
Given ﬁxed K, O, for every x ∈ K there is some r
x
such that B
rx
(x) ⊆ O
and [B
rx
(x)[ < α
−1
µ(B
rx
(x)). Since K is compact, we can choose a ﬁnite
subcover of K. Moreover, by Lemma A.29 we can reﬁne our set of balls such
that
[K[ ≤ 3
n
k
¸
i=1
[B
r
i
(x
i
)[ <
3
n
α
k
¸
i=1
µ(B
r
i
(x
i
)) ≤ 3
n
µ(O)
α
. (A.89)
To see the second claim, observe that
¦x ∈ A[ (Dµ)(x) > 0¦ =
∞
¸
j=1
¦x ∈ A[ (Dµ)(x) >
1
j
¦ (A.90)
and by the ﬁrst part [¦x ∈ A[ (Dµ)(x) >
1
j
¦[ = 0 for any j if µ(A) = 0.
232 A. Almost everything about Lebesgue integration
Theorem A.31 (Lebesgue). Let f be (locally) integrable, then for a.e. x ∈
R
n
we have
lim
r↓0
1
[B
r
(x)[
Br(x)
[f(y) −f(x)[dy = 0. (A.91)
Proof. Decompose f as f = g + h, where g is continuous and h
1
< ε
(Theorem 0.31) and abbreviate
D
r
(f)(x) =
1
[B
r
(x)[
Br(x)
[f(y) −f(x)[dy. (A.92)
Then, since limF
r
(g)(x) = 0 (for every x) and D
r
(f) ≤ D
r
(g) + D
r
(h) we
have
limsup
r↓0
D
r
(f) ≤ limsup
r↓0
D
r
(h) ≤ (Dµ)(x) +[h(x)[, (A.93)
where dµ = [h[dx. Using [¦x[ [h(x)[ ≥ α¦ ≤ α
−1
h
1
and the ﬁrst part of
Lemma A.30 we see
[¦x[ limsup
r↓0
D
r
(f)(x) ≥ α¦[ ≤ (3
n
+ 1)
ε
α
(A.94)
Since ε is arbitrary, the Lebesgue measure of this set must be zero for every
α. That is, the set where the limsup is positive has Lebesgue measure
zero.
The points where (A.91) holds are called Lebesgue points of f.
Note that the balls can be replaced by more general sets: A sequence of
sets A
j
(x) is said to shrink to x nicely if there are balls B
r
j
(x) with r
j
→ 0
and a constant ε > 0 such that A
j
(x) ⊆ B
r
j
(x) and [A
j
[ ≥ ε[B
r
j
(x)[. For
example A
j
(x) could be some balls or cubes (not necessarily containing x).
However, the portion of B
r
j
(x) which they occupy must not go to zero! For
example the rectangles (0,
1
j
) (0,
2
j
) ⊂ R
2
do shrink nicely to 0, but the
rectangles (0,
1
j
) (0,
2
j
2
) don’t.
Lemma A.32. Let f be (locally) integrable, then at every Lebesgue point
we have
f(x) = lim
j→∞
1
[A
j
(x)[
A
j
(x)
f(y)dy. (A.95)
whenever A
j
(x) shrinks to x nicely.
Proof. Let x be a Lebesgue point and choose some nicely shrinking sets
A
j
(x) with corresponding B
r
j
(x) and ε. Then
1
[A
j
(x)[
A
j
(x)
[f(y) −f(x)[dy ≤
1
ε[B
r
j
(x)[
Br
j
(x)
[f(y) −f(x)[dy (A.96)
and the claim follows.
A.7. Derivatives of measures 233
Corollary A.33. Suppose µ is an absolutely continuous Borel measure on
R, then its distribution function is diﬀerentiable a.e. and dµ(x) = µ
(x)dx.
As another consequence we obtain
Theorem A.34. Let µ be a Borel measure on R
n
. The derivative Dµ
exists a.e. with respect to Lebesgue measure and equals the RadonNikodym
derivative of the absolutely continuous part of µ with respect to Lebesgue
measure, that is,
µ
ac
(A) =
A
(Dµ)(x)dx. (A.97)
Proof. If dµ = f dx is absolutely continuous with respect to Lebesgue mea
sure the claim follows from Theorem A.31. To see the general case use the
Lebesgue decomposition of µ and let N be a support for the singular part
with [N[ = 0. Then (Dµ
sing
)(x) = 0 for a.e. x ∈ N
by the second part of
Lemma A.30.
In particular, µ is singular with respect to Lebesgue measure if and only
if Dµ = 0 a.e. with respect to Lebesgue measure.
Using the upper and lower derivatives we can also give supports for the
absolutely and singularly continuous parts.
Theorem A.35. The set ¦x[(Dµ)(x) = ∞¦ is a support for the singular
and ¦x[0 < (Dµ)(x) < ∞¦ is a support for the absolutely continuous part.
Proof. Suppose µ is purely singular ﬁrst. Let us show that the set O
k
=
¦x[ (Dµ)(x) < k¦ satisﬁes µ(O
k
) = 0 for every k ∈ N.
Let K ⊂ O
k
be compact, let V
j
⊃ K be some open set such that
[V
j
`K[ ≤
1
j
. For every x ∈ K there is some ε = ε(x) such that B
ε
(x) ⊆ V
j
and µ(B
ε
(x)) ≤ k[B
ε
(x)[. By compactness, ﬁnitely many of these balls cover
K and hence
µ(K) ≤
¸
i
µ(B
ε
i
(x
i
)) ≤ k
¸
i
[B
ε
i
(x
i
)[. (A.98)
Selecting disjoint balls us in Lemma A.29 further shows
µ(K) ≤ k3
n
¸
[B
ε
i
(x
i
)[ ≤ k3
n
[V
j
[. (A.99)
Letting j → ∞ we see µ(K) ≤ k3
n
[K[ and by regularity we even have
µ(A) ≤ k3
n
[A[ for every A ⊆ O
k
. Hence µ is absolutely continuous on O
k
and since we assumed µ to be singular we must have µ(O
k
) = 0.
Thus (Dµ
sing
)(x) = ∞ for a.e. x with respect to µ
sing
and we are done.
234 A. Almost everything about Lebesgue integration
Example. The Cantor function is constructed as follows: Take the sets
C
n
used in the construction of the Cantor set C: C
n
is the union of 2
n
closed intervals with 2
n
− 1 open gaps in between. Set f
n
equal to j/2
n
on the j’th gap of C
n
and extend it to [0, 1] by linear interpolation. Note
that, since we are creating precisely one new gap between every old gap
when going from C
n
to C
n+1
, the value of f
n+1
is the same as the value of
f
n
on the gaps of C
n
. In particular, f
n
− f
m

∞
≤ 2
−min(n,m)
and hence
we can deﬁne the Cantor function as f = lim
n→∞
f
n
. By construction f
is a continuous function which is constant on every subinterval of [0, 1]`C.
Since C is of Lebesgue measure zero, this set is of full Lebesgue measure
and hence f
= 0 a.e. in [0, 1. In particular, the corresponding measure, the
Cantor measure, is supported on C and purely singular with respect to
Lebesgue measure.
Bibliography
[1] W. O. Amrein, J. M. Jauch, and K. B. Sinha, Scattering Theory in Quantum
Mechanics, W. A. Benajmin Inc., New York, 1977
[2] H. Bauer, Measure and Integration Theory, de Gruyter, Berlin, 2001.
[3] H. L. Cycon, R. G. Froese, W. Kirsch and B. Simon, Schr¨odinger Operators,
Springer, Berlin, 1987.
[4] V. Enss, Asymptotic completeness for Quantum mechanical potential scattering,
Comm. Math. Phys. 61, 285291 (1978).
[5] V. Enß, Schr¨odinger Operators, lecture notes (private communication).
[6] I. Gohberg, S. Goldberg, and N. Krupnik, Traces and Determinants of Linear
Operators, Birkh¨auser, Basel, 2000.
[7] P. D. Hislop and I. M. Sigal, Introduction to Spectral Theory Springer, New York,
1996.
[8] J.L. Kelly, General Topology, Springer, New York, 1955.
[9] Yoram Last, Quantum dynamics and decompositions of singular continuous spec
tra, J. Funct. Anal. 142, 406445 (1996).
[10] E. Lieb and M. Loss, Analysis, American Mathematical Society, Providence,
1997.
[11] P. Perry, Mellin transforms and scattering Theory, Duke Math. J. 47, 187193
(1987).
[12] E. Prugoveˇcki, Quantum Mechanics in Hilbert Space, 2nd edition, Academic
Press, New York, 1981.
[13] M. Reed and B. Simon, Methods of Modern Mathematical Physics I. Functional
Analysis, rev. and enl. edition, Academic Press, San Diego, 1980.
[14] M. Reed and B. Simon, Methods of Modern Mathematical Physics II. Fourier
Analysis, SelfAdjointness, Academic Press, San Diego, 1975.
[15] M. Reed and B. Simon, Methods of Modern Mathematical Physics III. Scattering
Theory, Academic Press, San Diego, 1979.
[16] M. Reed and B. Simon, Methods of Modern Mathematical Physics IV. Analysis
of Operators, Academic Press, San Diego, 1978.
235
236 Bibliography
[17] J.R. Retherford, Hilbert Space: Compact Operators and the Trace Theorem, Cam
bridge UP, Cambridge, 1993.
[18] W. Rudin, Real and Complex Analysis, 3rd edition, McGrawHill, New York,
1987.
[19] M. Schechter, Operator Methods in Quantum Mechanics, North Holland, New
York, 1981.
[20] M. Schechter, Spectra of Partial Diﬀerential Operators, 2nd ed., North Holland,
Amsterdam, 1986.
[21] W. Thirring, Quantum Mechanics of Atoms and Molecules, Springer, New York,
1981.
[22] G. N. Watson, A Treatise on the Theory of Bessel Functions, 2nd ed., Cambridge
University Press, Cambridge, 1962.
[23] J. Weidmann, Linear Operators in Hilbert Spaces, Springer, New York, 1980.
[24] J. Weidmann, Lineare Operatoren in Hilbertr¨aumen, Teil 1: Grundlagen,
B.G.Teubner, Stuttgart, 2000.
[25] J. Weidmann, Lineare Operatoren in Hilbertr¨aumen, Teil 2: Anwendungen,
B.G.Teubner, Stuttgart, 2003.
Glossary of notations
AC(I) . . . absolutely continuous functions, 72
B = B
1
B
n
. . . Borel σﬁeld of R
n
, 210.
C(H) . . . set of compact operators, 112.
C(U) . . . set of continuous functions from U to C.
C
∞
(U) . . . set of functions in C(U) which vanish at ∞.
C(U, V ) . . . set of continuous functions from U to V .
C
∞
c
(U, V ) . . . set of compactly supported smooth functions
χ
Ω
(.) . . . characteristic function of the set Ω
dim . . . dimension of a linear space
dist(x, Y ) = inf
y∈Y
x −y, distance between x and Y
D(.) . . . domain of an operator
e . . . exponential function, e
z
= exp(z)
E(A) . . . expectation of an operator A, 47
T . . . Fourier transform, 139
H . . . Schr¨odinger operator, 177
H
0
. . . free Schr¨odinger operator, 142
H
m
(a, b) . . . Sobolev space, 72
H
m
(R
n
) . . . Sobolev space, 141
hull(.) . . . convex hull
H . . . a separable Hilbert space
i . . . complex unity, i
2
= −1
Im(.) . . . imaginary part of a complex number
inf . . . inﬁmum
Ker(A) . . . kernel of an operator A
L(X, Y ) . . . set of all bounded linear operators from X to Y , 20
237
238 Glossary of notations
L(X) = L(X, X)
L
p
(M, dµ) . . . Lebesgue space of p integrable functions, 22
L
p
loc
(M, dµ) . . . locally p integrable functions
L
p
c
(M, dµ) . . . compactly supported p integrable functions
L
∞
(M, dµ) . . . Lebesgue space of bounded functions, 23
L
∞
∞
(R
n
) . . . Lebesgue space of bounded functions vanishing at ∞
λ . . . a real number
max . . . maximum
´ . . . Mellin transform, 201
µ
ψ
. . . spectral measure, 82
N . . . the set of positive integers
N
0
= N ∪ ¦0¦
Ω . . . a Borel set
Ω
±
. . . wave operators, 197
P
A
(.) . . . family of spectral projections of an operator A
P
±
. . . projector onto outgoing/incoming states, 200
Q . . . the set of rational numbers
Q(.) . . . form domain of an operator, 84
R(I, X) . . . set of regulated functions, 98
R
A
(z) . . . resolvent of A, 62
Ran(A) . . . range of an operator A
rank(A) = dimRan(A), rank of an operator A, 111
Re(.) . . . real part of a complex number
ρ(A) . . . resolvent set of A, 61
R . . . the set of real numbers
S(I, X) . . . set of simple functions, 98
o(R
n
) . . . set of smooth functions with rapid decay, 139
sup . . . supremum
supp . . . support of a function
σ(A) . . . spectrum of an operator A, 61
σ
ac
(A) . . . absolutely continuous spectrum of A, 91
σ
sc
(A) . . . singular continuous spectrum of A, 91
σ
pp
(A) . . . pure point spectrum of A, 91
σ
p
(A) . . . point spectrum (set of eigenvalues) of A, 88
σ
d
(A) . . . discrete spectrum of A, 119
σ
ess
(A) . . . essential spectrum of A, 119
span(M) . . . set of ﬁnite linear combinations from M, 12
Z . . . the set of integers
z . . . a complex number
Glossary of notations 239
I . . . identity operator
√
z . . . square root of z with branch cut along (−∞, 0)
z
∗
. . . complex conjugation
A
∗
. . . adjoint of A, 51
A . . . closure of A, 55
ˆ
f = Tf, Fourier transform of f
ˇ
f = T
−1
f, inverse Fourier transform of f
. . . . norm in the Hilbert space H
.
p
. . . norm in the Banach space L
p
, 22
'., ..` . . . scalar product in H
E
ψ
(A) = 'ψ, Aψ` expectation value
∆
ψ
(A) = E
ψ
(A
2
) −E
ψ
(A)
2
variance
⊕ . . . orthogonal sum of linear spaces or operators, 38
∆ . . . Laplace operator, 142
∂ . . . gradient, 139
∂
α
. . . derivative, 139
M
⊥
. . . orthogonal complement, 36
(λ
1
, λ
2
) = ¦λ ∈ R[ λ
1
< λ < λ
2
¦, open interval
[λ
1
, λ
2
] = ¦λ ∈ R[ λ
1
≤ λ ≤ λ
2
¦, closed interval
ψ
n
→ ψ . . . norm convergence
ψ
n
ψ . . . weak convergence, 41
A
n
→ A . . . norm convergence
A
n
s
→ A . . . strong convergence, 43
A
n
A . . . weak convergence, 43
A
n
nr
→ A . . . norm resolvent convergence, 131
A
n
sr
→ A . . . strong resolvent convergence, 131
Index
a.e., see almost everywehre
Absolutely continuous
function, 72
measure, 227
Adjoint, 40
Algebra, 209
Almost everywhere, 212
Angular momentum operator, 151
Banach algebra, 21
Banach space, 11
Basis
orthonormal, 34
spectral, 80
Bessel function, 146
spherical, 185
Bessel inequality, 33
Borel
function, 219
measure, 211
set, 210
σalgebra, 210
Borel measure
regular, 211
Borel transform, 82
Creal, 71
Cantor function, 234
Cantor measure, 234
Cantor set, 212
CauchySchwarz inequality, 16
Caylay transform, 69
Ces`aro average, 110
Characteristic function, 220
Closed set, 5
Closure, 5
Commute, 101
Compact, 8
locally, 9
sequentially, 8
Complete, 6, 11
Conﬁguration space, 48
Conjugation, 71
Continuous, 6
Convolution, 140
Core, 55
Cover, 7
C
∗
algebra, 41
Cyclic vector, 80
Dense, 6
Dilation group, 179
Dirac measure, 211, 224
Dirichlet boundary condition, 162
Distance, 9
Distribution function, 211
Domain, 20, 48, 50
Eigenspace, 98
Eigenvalue, 61
multiplicity, 98
Eigenvector, 61
Element
adjoint, 41
normal, 41
positive, 41
selfadjoint, 41
unitary, 41
Essential supremum, 23
Expectation, 47
First resolvent formula, 63
241
242 Index
Form domain, 58, 84
Fourier series, 35
Fourier transform, 110, 139
Friedrichs extension, 61
Function
absolutely continuous, 72
Gaussian wave packet, 151
Gradient, 139
GramSchmidt orthogonalization, 35
Graph, 55
Graph norm, 55
Green’s function, 146
Ground state, 186
H¨older’s inequality, 23
Hamiltonian, 49
Harmonic oscillator, 154
Hausdorﬀ space, 5
Heisenberg picture, 114
Herglotz functions, 82
Hermite polynomials, 154
Hilbert space, 15, 31
separable, 35
Hydrogen atom, 178
Ideal, 41
Induced topology, 5
Inner product, 15
Inner product space, 15
Integrable, 222
Integral, 220
Interior, 5
Interior point, 4
Intertwining property, 198
Involution, 41
Ionisation, 192
Kernel, 20
l.c., see Limit circle
l.p., see Limit point
Laguerre polynomial, 185
associated, 185
Lebesgue measure, 212
Lebesgue point, 232
Legendre equation, 182
Lemma
RiemannLebesgue, 142
Lidiskij trace theorem, 127
Limit circle, 161
Limit point, 4, 161
Linear functional, 21, 37
Localization formula, 193
Meansquare deviation, 48
Measurable
function, 218
set, 210
Measure, 210
absolutely continuous, 227
complete, 218
ﬁnite, 210
growth point, 86
mutually singular, 227
product, 225
projectionvalued, 76
spectral, 82
support, 212
Measure space, 210
Mellin transform, 201
Metric space, 3
Minkowski’s inequality, 24
Molliﬁer, 26
Momentum operator, 150
Multiindex, 139
order, 139
Multiplicity
spectral, 81
Neighborhood, 4
Neumann function
spherical, 185
Neumann series, 64
Norm, 10
operator, 20
Norm resolvent convergence, 131
Normal, 10
Normalized, 15, 32
Normed space, 10
Nowhere dense, 27
Observable, 47
Oneparameter unitary group, 49
Open ball, 4
Open set, 4
Operator
adjoint, 51
bounded, 20
closable, 55
closed, 55
closure, 55
compact, 112
domain, 20, 50
ﬁnite rank, 111
Hermitian, 50
Hilbert–Schmidt, 122
linear, 20, 50
nonnegative, 58
normal, 79
positive, 58
relatively bounded, 117
relatively compact, 112
selfadjoint, 51
Index 243
semibounded, 61
strong convergence, 43
symmetric, 50
unitary, 33, 49
weak convergence, 43
Orthogonal, 15, 32
Orthogonal complement, 36
Orthogonal projection, 37
Orthogonal sum, 38
Outer measure, 216
Parallel, 15, 32
Parallelogram law, 17
Parseval’s identity, 141
Partial isometry, 85
Perpendicular, 15, 32
Phase space, 48
Pl¨ ucker identity, 161
Polar decomposition, 85
Polarization identity, 17, 33, 50
Position operator, 149
Positivity improving, 187
Positivity preserving, 187
Premeasure, 210
Probability density, 47
Product measure, 225
Projection, 41
Pythagorean theorem, 16
Quadratic form, 50
Range, 20
Rank, 111
Regulated function, 98
Relatively compact, 112
Resolution of the identity, 77
Resolvent, 62
Neumann series, 64
Resolvent set, 61
Riesz lemma, 37
Scalar product, 15
Scattering operator, 198
Scattering state, 198
Schatten pclass, 125
Schr¨odinger equation, 49
Second countable, 5
Second resolvent formula, 119
Selfadjoint
essentially, 55
Semimetric, 3
Separable, 6, 13
Short range, 203
σalgebra, 209
σﬁnite, 210
Simple function, 98, 220
Simple spectrum, 81
Singular values, 120
Span, 12
Spectral basis, 80
ordered, 90
Spectral measure
maximal, 89
Spectral theorem, 83
compact operators, 120
Spectral vector, 80
maximal, 89
Spectrum, 61
absolutely continuous, 91
discrete, 119
essential, 119
pure point, 91
singularly continuous, 91
Spherical harmonics, 183
∗algebra, 41
∗ideal, 41
Stieltjes inversion formula, 82
Stone’s formula, 100
Strong resolvent convergence, 131
SturmLiouville equation, 157
regular, 158
Subcover, 7
Subspace
reducing, 67
Superposition, 48
Tensor product, 39
Theorem
BanachSteinhaus, 28
closed graph, 57
dominated convergence, 223
Fubini, 226
HeineBorel, 9
HVZ, 192
Kato–Rellich, 118
Lebesgue decomposition, 229
monotone convergence, 221
Pythagorean, 32
RadonNikodym, 229
RAGE, 113
spectral, 84
spectral mapping, 90
Stone, 108
Stone–Weierstraß, 45
virial, 179
Weierstraß, 13
Weyl, 129
Wiener, 110
Topological space, 4
Topology
base, 5
product, 7
Total, 12
Trace, 127
244 Index
Trace class, 126
Triangel inequality, 11
Trotter product formula, 115
Uncertainty principle, 150
Uniform boundedness principle, 28
Unit vector, 15, 32
Unitary group
Generator, 49
Urysohn lemma, 10
Variance, 48
Vitali set, 212
Wave function, 47
Wave operators, 197
Weak convergence, 21, 41
Weierstraß approxiamation, 13
Weyl relation, 150
Weyl sequence, 64
singular, 128
WeylTitchmarsh mfunction, 172
Wronskian, 158
Gerald Teschl Fakult¨t f¨r Mathematik a u Nordbergstraße 15 Universit¨t Wien a 1090 Wien, Austria Email: Gerald.Teschl@univie.ac.at URL: http://www.mat.univie.ac.at/˜gerald/
2000 Mathematics subject classiﬁcation. 8101, 81Qxx, 4601
Abstract. This manuscript provides a selfcontained introduction to mathematical methods in quantum mechanics (spectral theory) with applications to Schr¨dinger operators. The ﬁrst part covers mathematical foundations o of quantum mechanics from selfadjointness, the spectral theorem, quantum dynamics (including Stone’s and the RAGE theorem) to perturbation theory for selfadjoint operators. The second part starts with a detailed study of the free Schr¨dinger opo erator respectively position, momentum and angular momentum operators. Then we develop WeylTitchmarsh theory for SturmLiouville operators and apply it to spherically symmetric problems, in particular to the hydrogen atom. Next we investigate selfadjointness of atomic Schr¨dinger operators o and their essential spectrum, in particular the HVZ theorem. Finally we have a look at scattering theory and prove asymptotic completeness in the short range case. Keywords and phrases. Schr¨dinger operators, quantum mechanics, uno bounded operators, spectral theory.
A Typeset by AMSL TEX and Makeindex. Version: April 19, 2006 Copyright c 19992005 by Gerald Teschl
Contents
Preface Part 0. Preliminaries Chapter 0. A ﬁrst look at Banach and Hilbert spaces §0.1. Warm up: Metric and topological spaces §0.2. The Banach space of continuous functions §0.3. The geometry of Hilbert spaces §0.4. Completeness §0.5. Bounded operators §0.6. Lebesgue Lp spaces §0.7. Appendix: The uniform boundedness principle Part 1. Mathematical Foundations of Quantum Mechanics Chapter 1. Hilbert spaces
vii
3 3 10 14 19 20 22 27
31 31 33 36 38 40 41 44 47 iii
§1.1. Hilbert spaces §1.2. Orthonormal bases §1.3. The projection theorem and the Riesz lemma §1.4. Orthogonal sums and tensor products §1.5. The C∗ algebra of bounded linear operators §1.6. Weak and strong convergence §1.7. Appendix: The Stone–Weierstraß theorem Chapter 2. Selfadjointness and spectrum
iv
Contents
§2.1. §2.2. §2.3. §2.4. §2.5. §2.6. Chapter §3.1. §3.2. §3.3. §3.4. Chapter §4.1. §4.2. §4.3. §4.4. §4.5. Chapter §5.1. §5.2. §5.3. Chapter §6.1. §6.2. §6.3. §6.4. §6.5.
Some quantum mechanics Selfadjoint operators Resolvents and spectra Orthogonal sums of operators Selfadjoint extensions Appendix: Absolutely continuous functions 3. The spectral theorem The spectral theorem More on Borel measures Spectral types Appendix: The Herglotz theorem 4. Applications of the spectral theorem Integral formulas Commuting operators The minmax theorem Estimating eigenspaces Tensor products of operators 5. Quantum dynamics The time evolution and Stone’s theorem The RAGE theorem The Trotter product formula
47 50 61 67 68 72 75 75 85 89 91 97 97 100 103 104 105 107 107 110 115
6. Perturbation theory for selfadjoint operators 117 Relatively bounded operators and the Kato–Rellich theorem 117 More on compact operators 119 Hilbert–Schmidt and trace class operators 122 Relatively compact operators and Weyl’s theorem 128 Strong and norm resolvent convergence 131
Part 2. Schr¨dinger Operators o Chapter §7.1. §7.2. §7.3. §7.4. 7. The free Schr¨dinger operator o The Fourier transform The free Schr¨dinger operator o The time evolution in the free case The resolvent and Green’s function 139 139 142 144 145
Abstract theory §12. Borel measures in a nut shell §A. Almost everything about Lebesgue integration §A.1.6. §9.1.4.2.3. §9. The eigenvalues of the hydrogen atom §10. Decomposition of measures §A.1. Atomic Schr¨dinger operators o §11.2. Measurable functions §A. Derivatives of measures Bibliography Glossary of notations Index 209 209 213 218 220 224 227 229 235 237 241 . Oneparticle Schr¨dinger operators o §10. limit point alternative Spectral transformations 149 149 151 154 157 157 161 168 177 177 178 181 184 186 189 189 191 197 197 200 202 Chapter 10. §8. Angular momentum §10.Contents v Chapter §8. Selfadjointness and spectrum §10. 8. Product measures §A. Incoming and outgoing states §12. The hydrogen atom §10. Chapter §9. Schr¨dinger operators with short range potentials o Part 3.3.1.3.5.4. One dimensional Schr¨dinger operators o SturmLiouville operators Weyl’s limit circle.2. §8. The HVZ theorem Chapter 12. Algebraic methods Position and momentum Angular momentum The harmonic oscillator 9. Nondegeneracy of the ground state Chapter 11.3.1.2.7.5. The Lebesgue integral §A.2. Extending a premasure to a measure §A. Appendix Appendix A.3.1. Scattering theory §12.2. Selfadjointness §11.
.
Preface Overview The present manuscript was written for my course Schr¨dinger Operators o held at the University of Vienna in Winter 1999. Moreover. The ﬁrst part is a stripped down introduction to spectral theory of unbounded operators where I try to introduce only those topics which are needed for the applications later on. Summer 2002. It is supposed to give a brief but rather self contained introduction to the mathematical methods of quantum mechanics with a view towards applications to Schr¨dinger operators. In addition. Hence I try to get to it as quickly as possible. I do not take the detour over bounded operators but I go straight for the unbounded case. This has the advantage that you will not get drowned in results which are never used again before you get to the applications. In particular. Nevertheless I still feel that the ﬁrst part should give you a solid background covering all important results which are usually taken for granted in more advanced books and research papers. and Summer 2005. I am not trying to provide an encyclopedic reference. existence of spectral measures is established via the Herglotz rather than the Riesz representation theorem since this approach paves the way for an investigation of spectral types via boundary values of the resolvent as the spectral parameter approaches the real line. My approach is built around the spectral theorem as the central object. vii . The applications presented are highly o selective and many important and interesting items are not touched.
In addition. However. provided you have the necessary background. In addition. Chapter 7. 3 are key chapters and you should study them in detail (except for Section 2. I compute the spectrum of the hydrogen atom. again I try to provide some mathematical details not found in physics textbooks. read Chapter 7. and Chapter 11. If you are interested in one dimensional models (SturmLiouville equations). In particular the RAGE theorem shows the connections between long time behavior and spectral types. For this reason there is a preliminary chapter reviewing all necessary results (including proofs). Chapter 5 contains two key results from quantum dynamics. Furthermore. which is a prerequisite for all others except for Chapter 9.viii Preface The second part starts with the free Schr¨dinger equation and computes o the free resolvent and time evolution. You should have a look at (e.g. I discuss position. while this assumption is reasonable for mathematics students. This is usually found in any physics textbook on quantum mechanics. momentum. Chapter 10. Chapter 6 is again of central importance and should be studied in detail. and angular momentum operators via algebraic methods. spectra of atoms (the HVZ theorem) and scattering theory. Hence. If you are interested in atoms.5 which can be skipped on ﬁrst reading). In particular. Prerequisites I assume some previous experience with Hilbert spaces and bounded linear operators which should be covered in any basic course on functional analysis. you can start reading in Chapter 1 or even Chapter 2. there is an appendix (again with proofs) providing all necessary results from measure theory. Chapter 4 should give you an idea of how the spectral theorem is used. The chapters in the second part are mostly independent of each others except for the ﬁrst one. Chapter 1 and Chapter 2. Chapters 2. with the only diﬀerence that I include some technical details which are usually not found there. Further topics are nondegeneracy of the ground state. Chapter 9 is all you need. Readers guide There is some intentional overlap between Chapter 0. Finally. you can skip the separation of variables (Sections 10.3 . it might not always be for physics students.) the ﬁrst section and you can come back to the remaining ones as needed. Stone’s theorem and the RAGE theorem.
the ﬁrst two sections of Chapter 10.ac.Preface ix and 10. and Chapter 12. If you are interested in scattering theory. Austria February.mat. and Karl Unterkoﬂer for pointing out errors in previous versions.4. Availability It is available from http://www. Maria HoﬀmannOstenhof. Zhenyou Huang. read Chapter 7. Chapter 5 is one of the key prerequisites in this case.univie.at/~gerald/ftp/bookschroe/ Acknowledgments I’d like to thank Volker Enß for making his lecture notes available to me and Wang Lanning. Harald Rindler. which require Chapter 9) method for computing the eigenvalues of the Hydrogen atom if you are happy with the fact that there are countably many which accumulate at the bottom of the continuous spectrum. 2005 . Gerald Teschl Vienna.
.
Part 0 Preliminaries .
.
z) ≤ d(x. Banach space. you can skip this entire chapter. or bounded linear operator. A metric space is a space X together with a function d : X × X → R such that (i) d(x. Warm up: Metric and topological spaces Before we begin I want to recall some basic facts from metric and topological spaces. y) = ( n xk −yk 2 )1/2 . y) = ( n (xk − yk )2 )1/2 k=1 is a metric space and so is Cn together with d(x. Euclidean space Rn together with d(x. 0. However. A crash course in measure theory can be found in the appendix. k=1 3 . you might want to at least browse through it to refresh your memory. y) + d(y. y) ≥ 0 (ii) d(x.1. I presume that you are familiar with these topics from your calculus course. For convenience.Chapter 0 A ﬁrst look at Banach and Hilbert spaces I assume that the reader has some basic familiarity with measure theory and functional analysis. some facts needed from Banach and Lp spaces are reviewed in this chapter. y) = d(y. Example. y) = 0 if and only if x = y (iii) d(x. z) (triangle inequality) If (ii) does not hold. x) (iv) d(x. A good reference is [8]. If you feel comfortable with terms like Lebesgue Lp spaces. d is called a semimetric.
Consider R with the usual metric and let U = (−1. limit point. Given two topologies O1 and O2 on X. y) = k=1 xk − yk  n n (0. A point x of some set U is called an interior point of U if U contains some ball around x.1) is called an open ball around x with radius r > 0.3) ˜ ˜ shows Br/√n ((x. If x is an interior point of U . A point x is called a limit point of U if Br (x) ∩ (U \{x}) = ∅ for every ball. where B. A set consisting only of interior points is called open. y)) ⊆ Br ((x. respectively. O is closed under ﬁnite intersections and arbitrary unions. There are usually diﬀerent choices for the topology. d. the open sets. satisfying (i)–(iii) is called a topological space.2) Since 1 √ n n xk  ≤ k=1 k=1 xk 2 ≤ k=1 xk  (0. X ∈ O (ii) O1 . we can equip Rn (or Cn ) with the Euclidean distance as before. The family of open sets O satisﬁes the following properties (i) ∅. For example.4 0. Then every point x ∈ U is an interior point of U . O2 ∈ O implies O1 ∩ O2 ∈ O (iii) {Oα } ⊆ O implies α Oα ∈O That is. Hence the topology is the same for both metrics. 1). then U is also called a neighborhood of x. or we could also use n ˜ d(x. In general. Moreover. y) < r} (0. The notions of interior point. Example. . B are balls com˜ puted using d. O1 is called weaker (or coarser) than O2 if and only if O1 ⊆ O2 . A ﬁrst look at Banach and Hilbert spaces The set Br (x) = {y ∈ Xd(x. and neighborhood carry over to topological spaces if we replace open ball by open set. X} and the discrete topology O = P(X) (the powerset of X). Two usually not very interesting examples are the trivial topology O = {∅. y)) ⊆ Br ((x. a space X together with a family of sets O. x is not a limit point of U if and only if it is an interior point of the complement of U . Note that diﬀerent metrics can give rise to the same topology. y)). The points ±1 are limit points of U . but U contains points arbitrarily close to x. Note that a limit point must not lie in U . Example.
y) ˜ d(x. If B ⊆ O is a base for the topology. O=O Example. are disjoint neighborhoods (a semimetric space will not be Hausdorﬀ). We can always replace a metric d by the bounded metric d(x.1. where d = d(x. Since O = x∈O U (x) we have Lemma 0. By construction the open balls B1/n (x) are a base for the topology in a metric space. X ∈ C (ii) C1 . y) without changing the topology. Example. 1]. there is some set O ∈ B with x ∈ O ⊆ U . (0. then every open set can be written as a union of elements from B. 1] ⊆ R is not open in the topology of X = R. y) > 0. A family of open sets B ⊆ O is called a base for the topology if for each x and each neighborhood U (x). (0.6) .0. C2 ∈ C implies C1 ∪ C2 ∈ C (iii) {Cα } ⊆ C implies α Cα ∈C That is. but it is open in the incuded topology when considered as a subset of Y = [−1. closed sets are closed under ﬁnite unions and arbitrary intersections. A topological space is called Hausdorﬀ space if for two diﬀerent points there are always two disjoint neighborhoods. It follows from de Morgan’s rules that the family of closed sets C satisﬁes (i) ∅. Example.1.4) 1 + d(x. then X is called second countable. Warm up: Metric and topological spaces 5 Example. Every subspace Y of a topological space X becomes a topological space ˜ of its own if we call O ⊆ Y open if there is some open set O ⊆ X such that ˜ ∩ Y (induced topology).U ⊆C C.O⊆U O. The smallest closed set containing a given set U is called the closure U= C∈C. If there exists a countable base.5) and the largest open set contained in a given set U is called the interior U◦ = O∈O. The set (0. y) = (0. Any metric space is a Hausdorﬀ space: Given two diﬀerent points x and y the balls Bd/2 (x) and Bd/2 (y). In the case of Rn (or Cn ) it even suﬃces to take balls with rational center and hence Rn (and Cn ) are second countable. The complement of an open set is called a closed set.
for every ε > 0 there is some N ∈ N such that d(xn . Clearly the limit is unique if it exists (this is not true for a semimetric).m ∈ B1/m (xn ) ∩ Y for all (n. Let A = {xn }n∈N be a dense set in X. Example. . A metric space is called separable if it contains a countable dense set. We write limn→∞ xn = x as usual in this case. Hence (nk . A ﬁrst look at Banach and Hilbert spaces It is straightforward to check that Lemma 0. In a Hausdorﬀ space the limit is unique. Every subset of X is again separable.m)∈J ⊆ Y is countable. f (y)) ≤ ε if dX (x. Let X be a topological space. Lemma 0. A sequence (xn )∞ ⊆ X is said to converge to some point x ∈ X if n=1 d(x.k . xn ) → 0. A closed subset of a complete metric space is again a complete metric space. Proof. A function between metric spaces X and Y is called continuous at a point x ∈ X if for every ε > 0 we can ﬁnd a δ > 0 such that dY (f (x). Note that convergence can also be equivalently formulated in terms of topological terms: A sequence xn converges to x if and only if for every neighborhood U of x there is some N ∈ N such that xn ∈ U for n ≥ N .7) If the converse is also true. Then there is some sequence xnk with d(xnk . To see that B is dense choose y ∈ Y . (0.k . m) ∈ J. Both Rn and Cn are complete metric spaces. A point x is clearly a limit point of U if and only if there is some sequence xn ∈ U converging to x. Then B = {yn. A set U is called dense. if its closure is all of X. Let X be a separable metric space. y) ≤ d(ynk . some elements of A must be at least arbitrarily close: Let J ⊆ N2 be the set of all pairs (n. xm ) ≤ ε n. that is if U = X. k) ∈ J and d(ynk .2. that is. The only problem is that A ∩ Y might contain no elements at all. Every convergent sequence is a Cauchy sequence.m }(n.6 0. m) for which B1/m (xn ) ∩ Y = ∅ and choose some yn. then the interior of U is the set of all interior points of U and the closure of U is the set of all limit points of U . that is. then X is called complete.3.8) If f is continuous at every point it is called continuous. Hence Lemma 0. xnk ) + d(xnk . if every Cauchy sequence has a limit. y) < 1/4. y) ≤ 2/k → 0. y) < δ. However. (0. m ≥ N.4.
the topology (and thus also convergence/continuity) is independent of this choice.10) As noted in our previous example. y2 )) = dX (x1 . If X and Y are just topological spaces. (0. x2 )2 + dY (y1 . Hence we can choose a sequence xn ∈ B1/n (x) such that f (xn ) ∈ f −1 (V ). x) + d(yn . Let X be a metric space. A cover of a set Y ⊆ X is a family of sets {Uα } such that Y ⊆ α Uα . Warm up: Metric and topological spaces 7 Lemma 0. y) ≤ d(xn . In particular. y2 )) = dX (x1 . by d(xn . A sequence (xn . Example. where the index set N is replaced by arbitrary directed sets. A cover is call open if all Uα are open. (x2 . In the case of metric spaces this clearly agrees with the topology deﬁned via the product metric (0. (x2 .5. y) ∈ O there are open neighborhoods U of x and V of y such that U × V ⊆ O.9) is a metric space. y2 ) (0. y) we see that d : X × X → R is continuous. y1 ). y2 )2 . Proof. If X and X are metric spaces then X × Y together with d((x1 . the product topology is deﬁned by calling O ⊆ X × Y open if for every point (x. Note: In a topological space.9). (iii) ⇒ (i): Choose V = Bε (f (x)) and observe that by (iii) Bδ (x) ⊆ f −1 (V ) for some δ. y) → x respectively onto the second (x. the projections onto the ﬁrst (x.e. y) → y coordinate are continuous. In particular. yn ) − d(x. Thus xn → x but f (xn ) → f (x).9) as follows: ˜ d((x1 . The following are equivalent (i) f is continuous at x (i. (ii) f (xn ) → f (x) whenever xn → x (iii) For every neighborhood V of f (x). x2 ) + dY (y1 . The last item implies that f is continuous if and only if the inverse image of every open (closed) set is again open (closed). A subset of {Uα } is called a subcover. (iii) is used as deﬁnition for continuity.8) holds).1. in general (ii) and (iii) will no longer be equivalent unless one uses generalized sequences.0. y1 ). so called nets. f −1 (V ) is a neighborhood of x. . (ii) ⇒ (iii): If (iii) does not hold there is a neighborhood V of f (x) such that Bδ (x) ⊆ f −1 (V ) for every δ.11) (0. yn ) converges to (x. y) if and only if xn → x and yn → y. (i) ⇒ (ii) is obvious. However. If we consider R × R we do not get the Euclidean distance of R2 unless we modify (0. (0.
A topological space is compact if and only if it has the ﬁnite intersection property: The intersection of a family of closed sets is empty if and only if the intersection of some ﬁnite subfamily is empty. then {f −1 (Oα )} is one for Y . . Fix x ∈ X\Y (if Y = X there is nothing to do). U (xj ) × V (xj . the open sets are a cover if and only if the corresponding closed sets have empty intersection. y)}y∈Y is an open cover of Y . j=1 (iv) Let {Oα } be an open cover for X × Y . for every y ∈ Y there are disjoint neighborhoods V (y) of y and Uy (x) of x. yk (xj )) ⊆ Oα(xj . For every n . Proof. (ii) Every closed subset of a compact set is compact. Lemma 0. Let X be a topological space. A subset K ⊂ X is called sequentially compact if every sequence has a convergent subsequence. y) × V (x. y) ∈ X × Y there is some α(x. y) ⊆ Oα(x. . (i) The continuous image of a compact set is compact.y) . yk (x)).y) . (iii) Let Y ⊆ X be compact. {U (x)}x∈X is an open cover and there are ﬁnitely many points xj such U (xj ) cover X.8 0. Moreover. We show that X\Y is open. By the deﬁnition of Hausdorﬀ. Hence for ﬁxed x.yk (xj )) cover X × Y . there are y1 . (v) A compact set is also sequentially compact. yn such that V (yj ) cover Y . y) such that (x. {V (x. A ﬁrst look at Banach and Hilbert spaces A subset K ⊂ X is called compact if every open cover has a ﬁnite subcover. By taking complements. Set U (x) = k U (x. Proof. (v) Let xn be a sequence which has no convergent subsequence. yk (x)) cover Y . Then {Oα } ∪ {X\Y } is an open cover for X. (iii) If X is Hausdorﬀ. . By compactness of Y . By construction. By deﬁnition of the product topology there is some open rectangle U (x. Then K = {xn } has no limit points and is hence compact by (ii). any compact set is closed. But then U (x) = n Uyj (x) is a neighborhood of x which does not intersect Y .7. y) ∈ Oα(x. For every (x. (i) Just observe that if {Oα } is an open cover for f (Y ). Hence there are ﬁnitely many points yk (x) such V (x. Since ﬁnite intersections of open sets are open. (ii) Let {Oα } be an open cover for the closed subset Y .6. to every family of open sets there is a corresponding family of closed sets and vice versa. Lemma 0. (iv) The product of compact sets is compact.
y). Proceeding like this we obtain a Cauchy sequence yn (note that by construction In+1 ⊆ In and hence yn − ym  ≤ b−a for m ≥ n). In a metric space compact and sequentially compact are equivalent. m=1 In particular. for every ε > 0 there must be an x such that Bε (x) ⊆ Oα for every α. we are done if we can show that for every open cover {Oα } there is some ε > 0 such that for every x we have Bε (x) ⊆ Oα for some α = α(x). Let y1 = xn1 be the ﬁrst one. Proof. Moreover. By Lemma 0. a+b ] ∪ [ a+b ]. Choose 1 ε = n and pick a corresponding xn . call it I1 .0.1. we k=1 have that Oα(xk ) is a cover as well. So it remains to show that there is such an ε. But choosing 1 ε ε n so large that n < 2 and d(xn . then x lies in some Oα and hence Bε (x) ⊆ Oα . Since X is sequentially compact. Then at least one of these two 2 2 intervals. Indeed. a contradiction. Example. Let xn be our sequence and divide I = [a. In Rn (or Cn ) a set is compact if and only if it is bounded and closed. by Lemma 0. b] has a convergent subsequence.8. n A topological space is called locally compact if every point has a compact neighborhood. Let X be a metric space. Please also recall the HeineBorel theorem: Theorem 0. However.9 (HeineBorel). First of all note that every cover of open balls with ﬁxed radius ε > 0 has a ﬁnite subcover. Proof.12) . x) < 2 we have B1/n (xn ) ⊆ Bε (x) ⊆ Oα contradicting our assumption. Y ) = inf d(x. ﬁnitely many suﬃce to cover K. y∈Y (0. with n2 > n1 as before. Then a subset is compact if and only if it is sequentially compact. If there were none. Subdivide I1 and pick y2 = xn2 . The distance between a point x ∈ X and a subset Y ⊆ X is dist(x. choosing {xk }n such that Bε (xk ) is a cover.8 it suﬃces to show that every sequence in I = [a. xm ) > ε for m < n. Since if this were false we could construct a sequence xn ∈ X\ n−1 Bε (xm ) such that d(xn . contains inﬁnitely many elements of our sequence. Lemma 0. Let x = lim xn . Rn is locally compact. Warm up: Metric and topological spaces 9 there is a ball Bεn (xn ) which contains only ﬁnitely many elements of K.7 (ii) and (iii) it suﬃces to show that a closed interval in I ⊆ R is compact. it is no restriction to assume xn converges (after maybe passing to a subsequence).
0. In particular. 1] such that f is zero on C1 and one on C2 . Y ) − dist(z. one can choose f with compact support. If X is locally compact and U is compact.11 (Urysohn). Lemma 0. 1]). Now replace C2 by X\O.10. then dist(x. Note that Urysohn’s lemma implies that a metric space is normal. z). Proof. Suppose C1 and C2 are disjoint closed subsets of a metric space X. Y ) ≤ dist(x. . Since U is compact. is the maximum norm: f (x) − g(x) ∞ = max f (x) − g(x).10 0. z). that is.14) x∈I It is not hard to see that with this deﬁnition C(I) becomes a normed linear space: A normed linear space X is a vector space X over C (or R) with a realvalued function (the norm) . such that • f ≥ 0 for all f ∈ X and f = 0 if and only if f = 0. Y ) ≤ d(x. x → dist(x. for every x. z) + d(z.13) For the second claim. choose f as in Urysohn’s lemma and set O1 = f −1 ([0. Lemma 0. there is a ball Bε (x) such that Bε (x) is compact and Bε (x) ⊂ X\C2 . Y ) ≤ dist(x. ﬁnitely many of them cover C1 and we can choose the union of those balls to be O. Interchanging x and z shows dist(z. To prove the ﬁrst claim set f (x) = dist(x.C2 ) dist(x. Y ) − dist(x. A ﬁrst look at Banach and Hilbert spaces Note that x ∈ Y if and only if dist(x. In fact.C1 )+dist(x. In fact. z)+dist(z. z). 1/2)) respectively O2 = f −1 ((1/2. we will always consider complex valued functions! One way of declaring a distance.2. The Banach space of continuous functions Now let us have a ﬁrst look at Banach spaces by investigating set of continuous functions C(I) on a compact interval I = [a. Y ). j = 1.C2 ) . Y ) = 0. for any two disjoint closed sets C1 and C2 . observe that there is an open set O such that O is compact and C1 ⊂ O ⊂ O ⊂ X\C2 . 2. b] ⊂ R. y) ≤ d(x. Proof. Taking the inﬁmum in the triangle inequality d(x. Since we want to handle complex models. Y ) ≤ dist(x. wellknown from calculus. Let X be a metric space. there are disjoint open sets O1 and O2 such that Cj ⊆ Oj . Hence dist(x. Y ) is continuous. y) shows dist(x. Then there is a continuous function f : X → [0. Y )−dist(z. (0. (0.
From the triangle inequality we also get the inverse triangle inequality (Problem 0. Let an = (an )∞ be j j=1 a Cauchy sequence.17) for any ﬁnite k. The space 1 (N) of all sequences a = (aj )∞ for which the norm j=1 ∞ a is ﬁnite. 1 satisﬁes the three requirements for a norm and (iii) 1 (N) is complete. First of all observe k k k aj + bj  ≤ j=1 j=1 aj  + j=1 bj  ≤ a 1 + b 1 (0.16) To show this. and • f + g ≤ f + g for all f. 1 = j=1 aj  (0. In fact. vector addition. This implies in particular am − an  ≤ ε for j j any ﬁxed j. g) = f −g and hence we know when a sequence of vectors fn converges to a vector f . for given ε > 0 we can ﬁnd an Nε such that am − an 1 ≤ ε for m. We will write fn → f or limn→∞ fn = f . (0. that is closed under addition and scalar multiplication (ii) .15) Once we have a norm. n ≥ Nε . that is.2). It remains to show that 1 (N) is complete. Moreover. The Banach space of continuous functions 11 • λ f = λ f for all λ ∈ C and f ∈ X. A complete normed space is called a Banach space. and multiplication by scalars are continuous (Problem 0. That 1 (N) is closed under scalar multiplication and the two other properties of a norm are straightforward. as usual. Example. is a Banach space. In addition to the concept of convergence we have also the concept of a Cauchy sequence and hence the concept of completeness: A normed space is called complete if every Cauchy sequence has a limit. it is not hard to see that the norm. a mapping F : X → Y between to normed spaces is called continuous if fn → f implies F (fn ) → F (f ).2.0. in this case. Thus an is a Cauchy sequence for ﬁxed j and by completeness j of C has a limit: limn→∞ an = aj .1)  f − g ≤ f −g . we have a distance d(f.18) . we need to verify three things: (i) 1 (N) is a Vector space. Letting k → ∞ we conclude that 1 (N) is closed under addition and that the triangle inequality holds. Now consider j k am − an  ≤ ε j j j=1 (0. g ∈ X (triangle inequality).
in the language of real analysis.22) cn un . by completeness of C. Now let us look at the case where fn is only a Cauchy sequence. whether it is continuous or not. we would even prefer a countable basis.24) then the span span{un } (the set of all ﬁnite linear combinations) of {un } is dense in X. Thus we get a limiting function f (x).23) that is. if there is a set {un } ⊂ X of linearly independent vectors such that every element f ∈ X can be written as f= n ∀m.12. that is. In particular. Moreover. n→∞ x∈I (0. j j=1 (0. Fortunately. in other words Theorem 0. fn converges uniformly to f . If such a basis exists. x ∈ I (0. Now what about convergence in this space? A sequence of functions fn (x) converges to f if and only if n→∞ lim f − fn = lim sup fn (x) − f (x) = 0. n > Nε . C(I) with the maximum norm is a Banach space. there is a wellknown result from real analysis which tells us that the uniform limit of continuous functions is again continuous. The space with the norm ∞ (N) of all bounded sequences a = (aj )∞ together j=1 a ∞ = sup aj  j∈N (0. n n n Example. (0. A set whose span is dense is called total and if we have a total set. Hence (a−an ) ∈ 1 (N) and since a ∈ 1 (N) we ﬁnally conclude a = a + (a − a ) ∈ 1 (N). that is. up to this point we don’t know whether it is in our vector space C(I) or not. In order to have only countable sums. A ﬁrst look at Banach and Hilbert spaces and take m → ∞: k aj − an  ≤ ε. there is a limit f (x) for each x.12 0.3). However. letting m → ∞ in fm (x) − fn (x) ≤ ε we see f (x) − fn (x) ≤ ε ∀n > Nε . fn (x) converges uniformly to f (x). (0. we also have a countable dense set (consider only linear combinations .19) Since this holds for any ﬁnite k we even have a−an 1 ≤ ε. Then fn (x) is clearly a Cauchy sequence of real numbers for any ﬁxed x ∈ I. x ∈ I. Next we want to know if there is a basis for C(I). Hence f (x) ∈ C(I) and thus every Cauchy sequence in C(I) converges.20) is a Banach space (Problem 0.21) That is. cn ∈ C. Or.
27) (Remark: The integral is known as Beta function and the asymptotics follow from Stirling’s formula.) Lemma 0.29) converges uniformly to f (x). The Banach space 1 (N) is separable. Example. The Banach space of continuous functions 13 with rational coeﬃcients – show this). 1 ] which vanishes at the endpoints. In fact. By considering f (x) − f (a) + (f (b) − f (a))(x − b) it is no loss to assume that f vanishes at the boundary points.14 (Smoothing). δ > 0.) 1 1 Then for every f ∈ C[− 2 . Moreover. 2 2 Now the claim follows from the lemma below using 1 un (x) = (1 − x2 )n . In where 1 n! (1 − x2 )n dx = 1 1 In = ( 2 + 1) · · · ( 1 + n) −1 2 2 = √ Γ(1 + n) π 3 = Γ( 2 + n) π 1 (1 + O( )). without restriction we only consider I = [ −1 . Then the set of polynomials is dense in C(I). then n = a k=1 ak δ ∞ a − an 1 = j=n+1 aj  → 0 (0. A normed linear space containing a countable dense set is called separable. Let I be a compact interval. the set of vectors n n δ n .26) (0.0. Proof. 1] such that un (x)dx = 1 x≤1 and δ≤x≤1 un (x)dx → 0. f (− 2 ) = 2 f ( 1 ) = 0. with δn = 1 and δm = 0.13 (Weierstraß).25) since an = aj for 1 ≤ j ≤ n and an = 0 for j > n. un has mass one and concentrates near x = 0 as n → ∞.2. . Let f (x) ∈ C(I) be given. 1 ] (why?). n = m is total: Let a ∈ 1 (N) be given and set n k . (0. Let un (x) be a sequence of nonnegative continuous functions on [−1.28) (In other words. n n (0. j j Luckily this is also the case for C(I): Theorem 0. we have that 2 1/2 fn (x) = −1/2 un (x − y)f (y)dy (0.
3. The geometry of Hilbert spaces So it looks like C(I) has all the properties we want. That is. for given ε we can ﬁnd a δ (independent of x) such that f (x)−f (y) ≤ ε whenever x−y ≤ δ.2. and λn gn → λg. Show that ∞ (N) is not separable (Hint: Consider sequences which take only the value one and zero. Problem 0. Show that the norm. which proves the claim. vector addition. C(I) is separable. if fn → f . Problem 0. Moreover.1. fn + gn → f + g. The same is true for 1 (N). and λn → λ then fn → f . Show that  f − g  ≤ f − g . and multiplication by scalars are continuous.14 0. The same idea is used to approximate noncontinuous functions by smooth ones (of course the convergence will no longer be uniform in this case). but not for ∞ (N) (Problem 0. gn → g. Corollary 0. Show that ∞ (N) is a Banach space. 0.30) In fact. Problem 0. either the distance of x to one of the boundary points ± 1 is smaller 2 than δ and hence f (x) ≤ ε or otherwise the diﬀerence between one and the integral is smaller than ε.3.4. How many are there? What is the distance between two such sequences?).x−y≥δ = ε + 2M ε + M ε = (1 + 3M )ε. However.4)! Problem 0. (0. hence the title smoothing lemma. there is still one thing missing: How should we deﬁne orthogonality in C(I)? In . Using this we have 1/2 fn (x) − f (x) ≤ −1/2 un (x − y)f (y) − f (x)dy + M ε un (x − y)f (y) − f (x)dy y≤1/2.x−y≤δ ≤ + un (x − y)f (y) − f (x)dy + M ε y≤1/2. (0.15. Now abbreviate M = max f and note 1/2 1/2 f (x)− −1/2 un (x−y)f (x)dy = f (x) 1− −1/2 un (x−y)dy ≤ M ε. Since f is uniformly continuous. A ﬁrst look at Banach and Hilbert spaces Proof. we can choose n such that δ≤y≤1 un (y)dy ≤ ε.31) Note that fn will be as smooth as un .
f . (0. (0. α1 g1 + α2 g2 ∗ ∗ = α1 f1 . g2 α1 . α2 ∈ C. two vectors are called orthogonal if their scalar product vanishes.35) with scalar product ∞ a. j (0. A somewhat more interesting example is the Hilbert space that is. A vector f ∈ H is called normalized or unit vector if f = 1. α1 f1 + α2 f2 . g ∈ H are called orthogonal or perpendicular (f ⊥ g) if f. f is indeed a norm.. .3. (aj )∞ j=1 j=1 aj 2 < ∞ (0. b = j=1 a∗ bj . f > 0 for f = 0 (ii) f. Clearly Cn with the usual scalar product n a. A map . f ∗ (positive deﬁnite) (symmetry) is called inner product or scalar product. . g f. Only the triangle inequality is nontrivial which will follow from the CauchySchwarz inequality below. The geometry of Hilbert spaces 15 Euclidean space.5) Of course I still owe you a proof for the claim that f. Example. A skew linear form satisfying the requirements (i) f. so we would need a scalar product: Suppose H is a vector space. g = g. ..0.33) The pair (H.34) is a (ﬁnite dimensional) Hilbert space. : H × H → C is called skew linear form if it is conjugate linear in the ﬁrst and linear in the second argument. If H is complete it is called a Hilbert space. = α1 f.. ) is called inner product space. b = j=1 a∗ bj j (0. that is. g = 0 and parallel if one is a multiple of the other.36) (Show that this is in fact a separable Hilbert space! Problem 0. the set of all sequences ∞ 2 (N). . Associated with every scalar product is a norm f = f. Example. g1 + α2 f. Two vectors f. g .32) where ‘∗’ denotes complex conjugation.. g + α2 f2 .
then for every f. f u (0. f u. f ⊥ g. (0. gn → f.41) with equality if and only if f and g are parallel. g  ≤ f g (0. f w f f f⊥ f f f I f I u Taking any other vector parallel to u it is easy to see f − αu 2 = f⊥ + (f − αu) 2 = f⊥ 2 +  u. f +g 2 = f 2 + f. f 2 + f⊥ 2 .42) But let us return to C(I). f u is the unique vector parallel to u which is closest to f. Let H0 be an inner product space. f⊥ = u. As a ﬁrst consequence we obtain the CauchySchwarzBunjakowski inequality: Theorem 0.40) and hence f = u. if fn → f and gn → g we have fn . But then the claim follows from f 2 =  g. f − u. is indeed a norm. g ∈ H0 we have  f. u = 0.38) and f⊥ deﬁned via f⊥ = f − u.37) which is one line of computation. f + g 2 ≤ ( f + g )2 . g . As another consequence we infer that the map . Can we ﬁnd a scalar product which has the maximum norm as associated norm? Unfortunately the answer is no! The . f u = u. A ﬁrst look at Banach and Hilbert spaces For two orthogonal vectors we have the Pythagorean theorem: f +g 2 = f 2 + g 2.16 (CauchySchwarzBunjakowski). Suppose u is a unit vector. then the projection of f in the direction of u is given by f = u. It suﬃces to prove the case g = 1. that is. f − u.16 0. (0. g + g. Proof. Note that the CauchySchwarz inequality entails that the scalar product is continuous in both variables.39) is perpendicular to u since u. f u (0. f − α2 (0.
3. The geometry of Hilbert spaces 17 reason is that the maximum norm does not satisfy the parallelogram law (Problem 0. If an inner product space is given. But how do we deﬁne a scalar product on C(I)? One possibility is b 2 + f −g 2 =2 f 2 +2 g 2 (0. g ) and 2 thus s(f. 0) = 0) shows s(f. veriﬁcation of the parallelogram law and the polarization identity is straight forward (Problem 0. λg) for every positive rational λ. g) + s(f. 1 if there is a constant m > 0 such that f 1 2 is ≤m f 2. By continuity (check this!) this holds for all λ > 0 and s(f. g = a f ∗ (x)g(x)dx. that is λs(f.46) 2 Now choosing h = 0 (and using s(f. A norm is associated with a scalar product if and only if the parallelogram law f +g holds. g) = s(g. h) = 2s(f. (0.17 (Jordanvon Neumann). Theorem 0. 2 on a space X. 2n g). Note that cont we have f ≤ b − a f ∞ (0. f )∗ are straightforward to check. we deﬁne 1 f + g 2 − f − g 2 + i f − ig 2 − i f + ig 2 . another straightforward computation using the parallelogram law shows g+h s(f. g) = s(f. g) = 2s(f.45) s(f. (0.6). g) = 4 Then s(f.7).44) f. (0. by induction we infer m m 2n s(f. (0. (0. g) + s(f. To show the converse.6). 1 and . f ) = f 2 and s(f. ). g = 4 Proof. ig) = i s(f. In this case the scalar product can be recovered from its norm by virtue of the polarization identity 1 f + g 2 − f − g 2 + i f − ig 2 − i f + ig 2 . Then . said to be stronger than . g) ﬁnishes the proof. Furthermore. −g) = −s(f.47) The corresponding inner product space is denoted by L2 (I).0.43) f. cont Suppose we have two norms . Note that the parallelogram law and the polarization identity even hold for skew linear forms (Problem 0.48) and hence the maximum norm is stronger than the L2 norm. g + h).49) . g) = s(f. g) respectively s(f. h) = s(f. Moreover.
52) and we can choose m2 = j 1. 1 and . Now choose m1 = 1/m. j αj uj 2 = j αj  . 2 is 2 2 the usual Euclidean norm. 2] and deﬁne 1 0≤x≤1− n 0. . is also a . 1 1 + n(x − 1).8)! This shows that in inﬁnite dimensional spaces diﬀerent norms will give raise to diﬀerent convergent sequences! In fact. Theorem 0. 1≤x≤2 (0. If .51) m2 Proof. . In particular. . 1 Cauchy sequence. In particular.18. L2 is separable. then all norms are equivalent. then by the triangle and Cauchy Schwartz inequalities f 1 ≤ j αj  uj uj 1 ≤ j uj 2 1 f 2 (0. Take I = [0. the key to solving problems in inﬁnite dimensional spaces is often ﬁnding the right norm! This is something which cannot happen in the ﬁnite dimensional case. but there is no limit in L2 ! cont cont Clearly the limit should be the step function which is 0 for 0 ≤ x < 1 and 1 for 1 ≤ x ≤ 2. . Thus . for given two norms . 1 ) it is also continuos in (X.5. . 2 ) and if a set is dense in (X.18 0. 2 it is also convergent with respect to . But is it also complete? Unfortunately cont the answer is no: Example. Problem 0. but this step function is discontinuous (Problem 0. (0. then any . 1 is continuous with respect to . and assume that . 1 − n ≤ x ≤ 1 fn (x) = 1. Clearly we can choose a basis uj . That is. 2 Cauchy sequence Hence if a function F : X → Y is continuous in (X.19. If X is a ﬁnite dimensional case. 1 ). Show that 2 (N) is a separable Hilbert space.50) then fn (x) is a Cauchy sequence in L2 . 1 ≤ j ≤ n. Let f = j αj uj . 2 there are constants m1 and m2 such that 1 f 1 ≤ f 2 ≤ m1 f 1 . A ﬁrst look at Banach and Hilbert spaces It is straightforward to check that Lemma 0. 1 . 2 is stronger than . 2 ) it is also dense in (X. 1 . if fn is convergent with respect to . 2 and attains its minimum m > 0 on the unit sphere (which is compact by the HeineBorel theorem).
0. Consequently the norm of a Cauchy sequence (xn )∞ can be deﬁned by n=1 (xn )∞ n=1 = limn→∞ xn and is independent of the equivalence class (show ¯ ˜ this!). ˜ to see that X Lemma 0.4.54) 4 Problem 0. Call two Cauchy sequences equivalent if their diﬀerence con¯ verges to zero and denote by X the set of all equivalence classes. More precisely any other complete space which contains X as a dense subset is isomorphic to X.7. g) be a skew linear form and p(f ) = s(f.50).24 below). Completeness Since L2 cont is not complete. then xn converges. Moreover. 0. Let ξn = [(xn. This can for example be seen by showing that the identity map on X has a unique extension to X (compare Theorem 0.j )∞ ] j=1 ¯ be a Cauchy sequence in X. In particular it is no restriction to assume that a normed linear space or an inner product space is complete.53) and the polarization identity 1 s(f.4. However. Show that the maximum norm (on C[0. (0. .6. Completeness 19 Problem 0. Let me remark that the completion X is unique.j )∞ ] j=1 is its limit. It is easy ¯ (and X) inherit the vector space structure from X. If xn is a Cauchy sequence. deﬁned in (0.20. f ) the associated quadratic form. consider the set of all Cauchy ˜ sequences X. Then it is not hard to see that ξ = [(xj. (Outline) It remains to show that X is complete. g) = (p(f + g) − p(f − g) + i p(f − ig) − i p(f + ig)) . If X is a (incomplete) normed space. how can we obtain a Hilbert space out of it? Well the answer is simple: take the completion. Prove the parallelogram law p(f + g) + p(f − g) = 2p(f ) + 2p(g) (0. Problem 0. 1]) does not satisfy the parallelogram law.8.21. in the last example. Thus X is a normed space (X is not! why?). Prove the claims made about fn . Let s(f. X is a Banach space containing X as a dense subspace if we identify x ∈ X with the equivalence class of all sequences converging to x. ¯ Theorem 0. ¯ Proof. in the important case of L2 cont it is somewhat inconvenient to work with equivalence classes of Cauchy sequences and hence we will give a diﬀerent characterization using the Lebesgue integral later.
A ﬁrst look at Banach and Hilbert spaces 0. that is An → A.58) is indeed a norm is straightforward. it is no restriction to assume that it is deﬁned on all of X.57) are deﬁned as usual. If Y is complete and An is a Cauchy sequence of operators.58) f X =1 (0. Moreover.23. Furthermore. if A is bounded and densely deﬁned. then An f converges to an element g for every f . Proof. The set of all bounded linear operators from X to Y is denoted by L(X. Theorem 0. An operator A is bounded if and only if it is continuous.59) and hence continuous. It is a Banach space if Y is. If X = Y we write L(X. Suppose A is continuous but not bounded. . The kernel Ker(A) = {f ∈ D(A)Af = 0} and range Ran(A) = {Af f ∈ D(A)} = AD(A) (0.22. Bounded operators A linear map A between two normed spaces X and Y will be called a (linear) operator A : D(A) ⊆ X → Y. X) = L(X). Taking the limit m → ∞ we see An f − Af ≤ ε f .55) The linear subspace D(A) on which A is deﬁned. (0. The space L(X. Y ). That (0. Then fn = n un converges to 0 but Afn ≥ 1 does not converge to 0. By continuity of the vector operations. The converse is also true Theorem 0. Proof.58) is a normed space. Then there is a sequence 1 of unit vectors un such that Aun ≥ n. given ε > 0 there is some N such that An − Am ≤ ε for n. m ≥ N and thus An f − Am f ≤ ε f . By construction. The operator A is called bounded if the following operator norm A = sup Af Y (0. a bounded operator is Lipschitz continuous Af Y ≤ A f X (0. Y ) together with the operator norm (0.5.20 0. Deﬁne a new operator A via Af = g. is called the domain of A and is usually required to be dense. A is linear and by continuity of the norm Af = limn→∞ An f ≤ (limn→∞ An ) f it is bounded.56) is ﬁnite.
Let A ∈ L(X. d dx deﬁned on .24. 1] (max norm) and X = L2 (0. Show that the integral operator 1 A(B + C) = AB + BC. note that our multiplication is not commutative (unless X is one dimensional). C) is called a bounded linear functional and the space X ∗ = L(X. note that (0. (0. A. A sequence fn is said to converge weakly fn f if (fn ) → (f ) for every ∈ X ∗ . this extension can only be given by Af = lim Afn .60) To show that this deﬁnition is independent of the sequence fn → f . let gn → f be a second sequence and observe Afn − Agn = A(fn − gn ) ≤ A fn − gn → 0.64) Moreover. (0. If D(A) is dense. B. Y ) and let Y be a Banach space. deﬁned on D(K) = C[0. from continuity of the norm we conclude that the norm does not increase.10. 1] × [0. AB ≤ A B . Show that the diﬀerential operator A = D(A) = C 1 [0. y)f (y)dy. Since a bounded operator maps Cauchy sequences to Cauchy sequences. (0.5.65) where K(x. 1]). We even have an identity. y) ∈ C([0. 1] is a bounded operator both in X = C[0.0. A Banach space together with a multiplication satisfying the above requirements is called a Banach algebra. cont Problem 0.63) (0. An operator in L(X. 1] ⊂ C[0. Bounded operators 21 Theorem 0. f ∈ X.64) ensures that multiplication is continuous. Problem 0. which has the same norm. α (AB) = (αA)B = A (αB). n→∞ fn ∈ D(A). C) is called the dual space of X. it is easy to see that we have However. In particular. Finally. Proof.62) (Kf )(x) = 0 K(x. there is a unique (continuous) extension of A to X.61) From continuity of vector addition and scalar multiplication it follows that our extension is linear. (0. The Banach space of bounded linear operators L(X) even has a multiplication given by composition.9. 1). α ∈ C. Clearly this multiplication satisﬁes (A + B)C = AC + BC. and (AB)C = A(BC). the identity operator I satisfying I = 1. C ∈ L(X) (0. 1] is an unbounded operator.
dµ) is a linear space. there is a small technical problem (recall that a property is said to hold almost everywhere if the set where it fails to hold is contained in a set of measure zero): Lemma 0. Problem 0.25. A ﬁrst look at Banach and Hilbert spaces Problem 0. g)p ≤ 2p max(f p . Thus f p = 0 only implies f (x) = 0 for almost every x. then f p dµ = 0 X (0. dµ) the set of all complex valued measurable functions for which f p is ﬁnite. First of all note that Lp (X. Show that AB ≤ A B for every A. Proof. Lebesgue Lp spaces We ﬁx some measure space (X. Of course our hope is that Lp (X.69) . where An = 1 {x f (x) > n }. dµ). However.e.11. dµ) = Lp (X. Let f be measurable. (with respect to λ). Let λ be the Lebesgue measure on R. dµ).12. p is not a norm on Lp (X. µ) and deﬁne the Lp norm by 1/p f p = X f p dµ . (0. (0. B ∈ L(X).e.66) and denote by Lp (X. but not for all! Hence . dµ) is a Banach space. Let Θ be the Dirac measure centered at 0. Observe that we have A = {xf (x) = 0} = n An . then f (x) = 0 a. The converse is obvious. The way out of this misery is to identify functions which are equal almost everywhere: Let N (X. If f p dµ = 0 we must have µ(An ) = 0 for every n and hence µ(A) = limn→∞ µ(An ) = 0. 1≤p (0. dµ) and we can consider the quotient space Lp (X.6. dµ)/N (X.68) Then N (X. gp ) ≤ 2p (f p + gp ). since f + gp ≤ 2p max(f . Show that the multiplication in a Banach algebra X is continuous: xn → x and yn → y implies xn yn → xy. Note that the proof also shows that if f is not 0 almost everywhere. there is an ε > 0 such that µ({x f (x) ≥ ε}) > 0. dµ) is a linear subspace of Lp (X. Example. (with respect to Θ) if and only if f (0) = 0.67) if and only if f (x) = 0 almost everywhere with respect to µ. 0. Then the characteristic function of the rationals χQ is zero a. dµ) = {f f (x) = 0 µalmost everywhere}. Σ.22 0.
As before we set L∞ (X. which plays a central role in the theory of Lp spaces. and x = 0 is the only point which counts for Θ). The only problem is that if we want to identify functions equal almost everywhere. dµ) are strictly speaking equivalence classes of functions. in the case of Lebesgue measure). However. Observe that f p is well deﬁned on Lp (X. If Θ is the Dirac measure centered at 0.g.13. C is an essential bound if f (x) ≤ C almost everywhere and the essential supremum is the inﬁmum over all essential bounds. o 1 1 + =1 p q (0. The solution is the essential supremum f ∞ = inf{C  µ({x f (x) > C}) = 0}. we will still call them functions for notational convenience. Even though the elements of Lp (X. (0. o In particular.70) That is. the supremum is no longer independent of the equivalence class.6. dµ) then f g ∈ L1 (X. dµ) and fg 1 ≤ f p g q. As a preparation for proving that Lp is a Banach space. Theorem 0. If λ is the Lebesgue measure. which is just the triangle inequality for Lp . dµ) and observe that f ∞ (0. If you wonder where the ∞ comes from. dµ) turns out to be a Banach space.. it will imply Minkowski’s inequality. Let p and q be dual indices. dµ). that is.0. dµ) = B(X)/N (X. But before that let us also deﬁne L∞ (X. e. We will show this in the following sections. It should be the set of bounded measurable functions B(X) together with the sup norm. dµ). have a look at Problem 0. dµ) the value f (x) is not well deﬁned (unless there is a continuous representative and diﬀerent continuous functions are in diﬀerent equivalence classes.26 (H¨lder’s inequality).73) . With this modiﬁcation we are back in business since Lp (X. we will need H¨lder’s inequality. then the essential sup of χQ with respect to λ is 0.72) with 1 ≤ p ≤ ∞. If f ∈ Lp (X. Example. Lebesgue Lp spaces 23 If dµ is the Lebesgue measure on X ⊆ Rn we simply write Lp (X). note that for f ∈ Lp (X. dµ) and g ∈ Lq (X.71) is independent of the equivalence class. then the essential sup of χQ with respect to Θ is 1 (since χQ (0) = 1. (0.
we only consider 1 < p < ∞. dµ) is complete.28. Then ∞ G(x) = k=1 gk (x) (0. 2 Now consider gn = fn − fn−1 (set f0 = 0). This follows from n n gk  k=1 p ≤ k=1 gk (x) p ≤ f1 p + 1 2 (0. q = 1) follows directly from the properties of the integral and hence it remains to consider 1 < p. (0.80) . q = ∞ (respectively p = ∞. Hence we can drop some terms such that 1 (0. (0.14) 1 1 a1/p b1/q ≤ a + b. The case p = 1. First of all it is no restriction to assume f p = g the elementary inequality (Problem 0.24 0.78) fn+1 − fn p ≤ n . Suppose fn is a Cauchy sequence.75) and ﬁnishes the proof. using (0. Theorem 0. The space Lp (X. dµ) is a Banach space.74) 1 p f p dµ + X 1 q gq dµ = 1 X (0. a.27 (Minkowski’s inequality). A ﬁrst look at Banach and Hilbert spaces Proof.79) is in Lp . Since the cases p = 1. Let f. q < ∞. Then.76) Proof. then f +g p ≤ f p + g p. ∞ are straightforward. p q with a = f p and b = gq and integrating over X gives f gdµ ≤ X q = 1. Finally it remains to show that Lp (X. Proof. It suﬃces to show that some subsequence converges (show this). dµ) is a normed linear space. Using f +gp ≤ f  f +gp−1 +g f +gp−1 we obtain from H¨lder’s o inequality (note (p − 1)q = p) f +g p p ≤ f p p (f + g)p−1 q + g p (f + g)p−1 q = ( f + g p ) (f + g) p−1 p .77) This shows that Lp (X. b ≥ 0. g ∈ Lp (X. As a consequence we also get Theorem 0. dµ).
dµ).81) is absolutely convergent for those x. is total. Then.31. Suppose X is a second countable topological space (i. By monotone convergence. Proof. Proof. In particular. As in the previous proof the set of all characteristic functions χK (x) with K compact is total (using inner regularity). there is a decreasing sequence of open sets On such that µ(On ) → µ(A). Moreover. The set of all characteristic functions χA (x) with A ∈ Σ and µ(A) < ∞..6. If fn − f p → 0 then there is a subsequence which converges pointwise almost everywhere. it has a countable basis) and µ is a regular Borel measure. Finally let B be a countable basis for the topology. Since f (x) − fn (x)p converges to zero almost everywhere and f (x) − fn (x)p ≤ 2p G(x)p ∈ L1 . χO with O ˜ To ﬁnish this chapter. Now our strategy is as follows: Using outer regularity we can restrict A to open sets and using the existence of a countable base. dµ). Since µ(A) < ∞ it is no restriction to assume µ(On ) < ∞. Now let f (x) be this limit. G(x) < ∞ almost everywhere and the sum ∞ gn (x) = lim fn (x) n=1 n→∞ (0. Lebesgue Lp spaces 25 using the monotone convergence theorem. Lemma 0.0. every open set O can be written as ˜ ˜ O = ∞ Oj with Oj ∈ B.30. in the proof of the last theorem we have seen: Corollary 0. It even turns out that Lp is separable. by considering the set of all ﬁnite j=1 ˜ unions of elements from B it is no restriction to assume n Oj ∈ B. Fix A. Theorem 0. is total by construction of the integral. we can restrict A to open sets from this base. let us show that continuous functions are dense in Lp . Thus the set of all characteristic functions χO (x) with O open and µ(O) < ∞. dominated convergence shows f − fn p → 0. By outer regularity. and thus µ(On \A) = µ(On ) − µ(A) → 0.e. 1 ≤ p < ∞. In particular. Then Lp (X. Hence j=1 ˜ ˜ there is an increasing sequence On O with On ∈ B. Hence it suﬃces to show .29. Then the set Cc (X) of continuous functions with compact support is dense in Lp (X. Now dominated convergence implies χA − χOn p → 0. χO − χOn p → 0 and hence the set of all characteristic functions ˜ ˜ ∈ B is total. 1 ≤ p < ∞ is separable. Let X be a locally compact metric space and let µ be a σﬁnite regular Borel measure.
Then for any (uniformly) continuous function f : Rn → C we have that fk (x) = Rn uk (x − y)f (y)dy (0. then the set Cc (X) of p (X.11) there is a continuous function fε which is one on K and 0 outside O. Let u be a molliﬁer in Rn and set uk (x) = k n u(k x).17): Lemma 0.82) we have fε − χK → 0 and we are done. By outer regularity there is an open set O ⊃ K such that µ(O\K) ≤ ε. Now we are ready to prove ∞ Theorem 0. ∞ A nonnegative function u ∈ Cc (Rn ) is called a molliﬁer if u(x)dx = 1 Rn (0. By Urysohn’s lemma (Lemma 0. By our previous result it suﬃces to show that any continuous function f (x) with compact support can be approximated by smooth ones. Now choose a molliﬁer u and observe that fk has compact support (since f .84) is in C ∞ (Rn ) and converges to f (uniformly). dµ).26 0. If X is some subset of Rn we can do even better. A ﬁrst look at Banach and Hilbert spaces that χK (x) can be approximated by continuous functions. it is no restriction to assume X = Rn . all smooth functions with compact support is dense in L Proof. By setting f (x) = 0 for x ∈ X.83) 1 The standard molliﬁer is u(x) = exp( x2 −1 ) for x < 1 and u(x) = 0 else. If X ⊆ Rn and µ is a Borel measure. Since χK − fε p dµ = X O\K fε p dµ ≤ µ(O\K) ≤ ε (0. 1 ≤ p < ∞.33.32. If we scale a molliﬁer according to uk (x) = k n u(k x) such that its mass is preserved ( uk 1 = 1) and it concentrates more and more around the origin T u k E we have the following result (Problem 0.
Show that if f ∈ Lp1 ∩ Lp2 . Prove Lemma 0.74). Moreover.8. Theorem 0. Problem 0. (Hint: To show that fk is smooth use Problem A. since f has compact support it is uniformly continuous and fk → f uniformly.15. that is. Let 0 < θ < 1. We can assume that the sets Xn are closed n=1 and none of them contains a ball. (Hint: Show that f (x) = (1 − t) + tx − xt . 1]. (Hint: Start with the function f0 (x) = x−α which has a single pole at 0.) Problem 0. We will construct a Cauchy sequence xn which stays away from all Xn . Appendix: The uniform boundedness principle Recall that the interior of a set is the largest open subset (that is.18. Appendix: The uniform boundedness principle 27 has). Let X be a complete metric space.86) + 1 q = 1.16 (Lyapunov inequality).17. Prove (0. then fj (x) = f0 (x − xj ) has a pole at xj .87) = θ p1 + 1−θ p2 . X\Xn is open and nonempty for every n. (0. The key to several important theorems about Banach spaces is the observation that a Banach space cannot be the countable union of nowhere dense sets. . then X cannot be the countable union of nowhere dense sets. then f ∈ Lp and f where 1 p p ≤ f θ p1 f 1−θ p2 .) Problem 0.0. But this implies fk → f in Lp . 1) which has a pole at every rational number in [0.) 0. Show that p→∞ lim f p = f ∞ (0. A set is called nowhere dense if its closure has empty interior.14. Show the following generalization of H¨lder’s inequality o fg where 1 p r ≤ f p g q.7 and A. the union of all open subsets). Suppose µ(X) < ∞. 0 < t < 1 satisﬁes f (a/b) ≥ 0 = f (1).13. Construct a function f ∈ Lp (0.7.32.34 (Baire category theorem). x > 0. r Problem 0. Problem 0. Proof. Suppose X = ∞ Xn .7. Problem 0.85) for any bounded measurable function. (0.
Moreover.91) .35 (BanachSteinhaus). since X2 cannot contain Br1 (x1 ) there is some x2 ∈ Br1 (x1 ) that is not in X2 . (Sets which can be written as countable union of nowhere dense sets are called of ﬁrst category.) In other words. by continuity of Aα and the norm. Proof. each Xn is an intersection of closed sets and hence closed. we can even assume Br1 (x1 ) ⊆ X\X1 . Since by construction xn ∈ BrN (xN ) for n ≥ N . Hence the name category theorem. Now observe Aα y ≤ Aα (y + x0 ) + Aα x0 ≤ n + Aα x0 (0.89) then n Xn = X by assumption. Let Xn = {x Aα x ≤ n for all α} = α {x Aα x ≤ n}. Suppose Aα x ≤ C(x) is bounded for ﬁxed x ∈ X.90) x for y < ε. Reducing r1 a little. Proceeding by induction we obtain a sequence of balls such that (0. hence without loss of generality rn → 0. contradicting our assumption that the Xn cover X. at least one Xn contains a ball of radius ε > 0. Setting y = ε x we obtain Aα x ≤ for any x. then Aα ≤ C is uniformly bounded. Let {Aα } ⊆ L(X. Now we come to the ﬁrst important consequence. if Xn ⊆ X is a sequence of closed subsets which cover X. the uniform boundedness principle. But x ∈ Brn (xn ) ⊆ X\Xn for every n. Since Br1 (x1 ) ∩ (X\X2 ) is open there is a closed ball Br2 (x2 ) ⊆ Br1 (x1 ) ∩ (X\X2 ). By Baire’s theorem at least one contains a ball of positive radius: Bε (x0 ) ⊂ Xn . (0. Moreover. we conclude that xn is Cauchy and converges to some point x ∈ X.28 0. n + C(x0 ) x ε (0. A ﬁrst look at Banach and Hilbert spaces Since X\X1 is open and nonempty there is a closed ball Br1 (x1 ) ⊆ X\X1 . Theorem 0. Y ) be a family of bounded operators.88) Brn (xn ) ⊆ Brn−1 (xn−1 ) ∩ (X\Xn ). All other sets are second category. Now observe that in every step we can choose rn as small as we please. Let X be a Banach space and Y some normed linear space.
Part 1 Mathematical Foundations of Quantum Mechanics .
.
ϕ  ≤ ψ ϕ (1. g = M f (x)∗ g(x)dµ(x). it is called a Hilbert space. ψ . A map . In quantum mechanics the phase space is always a Hilbert space H.1) The triangle inequality follows from the CauchySchwarzBunjakowski inequality:  ψ.. It is no restriction to assume that H is complete since one can easily replace it by its completion. A positive deﬁnite skew linear form is called inner product or scalar product. Associated with every scalar product is a norm ψ = ψ. . (1. dµ) is a Hilbert space with scalar product given by f.Chapter 1 Hilbert spaces The phase space in classical mechanics is the Euclidean space R2n (for the n position and n momentum coordinates). Hence the geometry of Hilbert spaces stands at the outset of our investigations. : H × H → C is called skew linear form if it is conjugate linear in the ﬁrst and linear in the second argument. The space L2 (M.. If H is complete with respect to the above norm.2) with equality if and only if ψ and ϕ are parallel. 1.3) 31 . Example. Hilbert spaces Suppose H is a vector space.1. (1.
In other words. Suppose {ϕj }n is an orthonormal set.8) where ψ and ψ⊥ are orthogonal. (1. ψ ϕj . Then every ψ ∈ H j=0 can be written as n (1. the set of all square summable sequences with scalar product f. ψ ϕ is perpendicular to ϕ. ψ ⊥ ϕ. ϕ = 0 and parallel if one is a multiple of the other.6) (1. then the projection of ψ in the direction of ϕ is given by ψ = ϕ. Two vectors ψ. Suppose ϕ is a unit vector.4) (Note that the second example is a special case of the ﬁrst one. (1. If ψ and ϕ are orthogonal we have the Pythagorean theorem: ψ+ϕ 2 = ψ 2 + ϕ 2. ϕj = 1. ϕj . (1.1. take M = R and µ a sum of Dirac measures. A set of vectors {ϕj } is called orthonormal set if ϕj . ψ ϕ and ψ⊥ deﬁned via ψ⊥ = ψ − ϕ.) A vector ψ ∈ H is called normalized or unit vector if ψ = 1.5) which is straightforward to check. n ψ 2 = j=0  ϕj . 2 (N) is a Hilbert space (1. g = j∈N ∗ fj gj . j=0 . ψ 2 + ψ⊥ 2 . These results can also be generalized to more than one vector. ϕ ∈ H are called orthogonal or perpendicular (ψ ⊥ ϕ) if ψ. every ψ in the span of {ϕj }n satisﬁes j=0 ˆ ψ − ψ ≥ ψ⊥ (1. Lemma 1. ψ = j=0 ϕj . ψ is uniquely characterized as the vector in the span of {ϕj }n being closest to ψ.9) ˆ Moreover. Hilbert spaces Similarly.10) ˆ with equality holding if and only if ψ = ψ . Moreover.7) ψ = ψ + ψ⊥ .32 1. In particular. ψ⊥ = 0 for all 1 ≤ j ≤ n. ϕk = 0 for j = k and ϕj .
ψ ∈ H1 . (a1 . H2 ) is called unitary if U preserves scalar products: U ϕ. U ψ 2 = ϕ. ψ − ψ = 0 and hence ψ and ψ⊥ = ψ − ψ are orthogonal. ϕ.2. (1. A straightforward calculation shows ϕj . a1 . Problem 1. a3 . (1.1. .14) 4 A bijective operator U ∈ L(H1 .15) By the polarization identity this is the case if and only if U preserves norms: U ψ 2 = ψ 1 for all ψ ∈ H1 . Then Bessel’s inequality (1. a2 .16) j∈J .5) iteratively. We start by assuming that J is countable. we need to generalize Lemma 1. j=0 Recall that a scalar product can be recovered from its norm by virtue of the polarization identity 1 ϕ. Now. ψ = ϕ + ψ 2 − ϕ − ψ 2 + i ϕ − iψ 2 − i ϕ + iψ 2 . . The formula for the norm follows by applying (1. (1.2. The two Hilbert space H1 and H2 are called unitarily equivalent in this case.11) in the span of {ϕj }n . . Orthonormal bases 33 Proof. since we cannot assume H to be a ﬁnite dimensional vector space. Then one computes j=0 ˆ ψ−ψ 2 = = ˆ ψ + ψ⊥ − ψ n 2 = ψ⊥ 2 ˆ + ψ −ψ 2 ψ⊥ 2 + j=0 cj − ϕj . The operator S : 2 (N) → 2 (N).9) we obtain Bessel’s inequality n  ϕj . Is it unitary? 1. ψ 1 . From (1. ) clearly satisﬁes U a = a .1. a2 .13) shows that  ϕj .1 to arbitrary orthonormal sets {ϕj }j∈J . ﬁx a vector n ˆ ψ= j=0 cj ϕj .13) with equality holding if and only if ψ lies in the span of {ϕj }n . Orthonormal bases Of course. ) → (0. . ψ 2 (1. .12) from which the last claim follows. ψ 2 ≤ ψ j=0 2 (1. ψ 2 (1.
18) j∈J is welldeﬁned and so is ϕj . For an orthonormal set {ϕj }j∈J the following conditions are equivalent: (i) {ϕj }j∈J is a maximal orthonormal set. Again.21) j∈J ˆ Moreover. An orthonormal set which is not a proper subset of any other orthonormal set is called an orthonormal basis due to following result: Theorem 1. Then every ψ ∈ H can be written as ψ = ψ + ψ⊥ . Thus it follows that  ϕj .23) . Hilbert spaces converges absolutely. In particular.22) ˆ with equality holding if and only if ψ = ψ . Now let J be arbitrary. ψ  ≥ ε. (ii) For every vector ψ ∈ H we have ψ= j∈J ϕj . ψ⊥ = 0 for all j ∈ J. ψ 2=  ϕj . Suppose {ϕj }j∈J is an orthonormal set. Theorem 1.34 1. ψ  is.19) In particular. Hence there are at most countably many j for which  ϕj . Moreover. ψ ϕj is Cauchy if and only 2 if j∈J  ϕj . ψ = j∈J ϕj . ψ ϕj . (1. ψ 2 (1. ψ 2 + ψ⊥ 2 . Bessel’s inequality shows that for any given ε > 0 there are at most ﬁnitely many j for which  ϕj . In other words. ψ ϕ j . ψ ϕj j∈K 2 = j∈K  ϕj . for any ﬁnite subset K ⊂ J we have ϕj . ψ 2 (1.17) by the Pythagorean theorem and thus j∈J ϕj .3. (1.2. ϕj . Moreover. ψ  > 0. ψ ϕj . (1. by continuity of the scalar product we see that Lemma 1. every ψ in the span of {ϕj }j∈J satisﬁes ˆ ψ − ψ ≥ ψ⊥ (1.1 holds for arbitrary orthonormal sets without modiﬁcations. ψ is uniquely characterized as the vector in the span of {ϕj }j∈J being closest to ψ.20) where ψ and ψ⊥ are orthogonal. Note that from Bessel’s inequality (which of course still holds) it follows that the map ψ → ψ is continuous. j∈J (1.
27) ψ 1 − ϕ0 . . (ii) ⇒ (iii): Follows since (ii) implies ψ⊥ = 0. But then {ϕj }j∈J ∪ {ψ⊥ } would be a larger orthonormal set. ψn . ϕj = 0 for all j ∈ J we conclude ψ 2 = 0 and hence ψ = 0. there would be a unit vector ϕ such that {ϕj }j∈J ∪ {ϕ} is larger orthonormal set. After throwing away some vectors we can assume that ψn+1 canj=0 not be expressed as a linear combinations of the vectors ψ0 .26) ψ0 Next we take ψ1 and remove the component parallel to ϕ0 and normalize again ψ 1 − ϕ0 . ˜ (i) ⇒ (ii): If ψ⊥ = 0 than we can normalize ψ⊥ to obtain a unit vector ψ⊥ ˜ which is orthogonal to all vectors ϕj . ψ n ϕj .2.28) This procedure is known as GramSchmidt orthogonalization. . In fact. (iv) ⇒ (i): If {ϕj }j∈J were not maximal. (iii) ⇒ (iv): If ψ. The set of functions 1 ϕn (x) = √ ein x . Now we can construct an orthonormal basis as follows: We begin by normalizing ψ0 ψ0 ϕ0 = . (Problem 1. Since ψ → ψ is continuous. Proof. if H is separable. (1.25) forms an orthonormal basis for H = L2 (0. Orthonormal bases 35 (iii) For every vector ψ ∈ H we have ψ 2 = j∈J  ϕj . contradicting maximality of {ϕj }j∈J .1. a contradiction. then there exists a countable total set {ψj }N . (1. ψ n ϕj ϕj . But ϕj . (1. ϕ = 0 for all j ∈ J implies ϕ = 0 by (iv). We will use the notation from Theorem 1. Hence we obtain an orthonormal set {ϕj }N such that span{ϕj }n = span{ψj }n j=0 j=0 j=0 . (1. (1. ψ = 0 for all j ∈ J implies ψ = 0. it suﬃces to check conditions (ii) and (iii) on a dense set. ψ 1 ϕ0 ϕ1 = .2. ψ 1 ϕ0 Proceeding like this we deﬁne recursively ϕn = ψn − ψn − n−1 j=0 n−1 j=0 ϕj .24) (iv) ϕj .17) A Hilbert space is separable if and only if there is a countable orthonormal basis. . Example. 2π n ∈ Z. ψ 2 . 2π). The corresponding orthogonal expansion is just the ordinary Fourier series.
it turns out that. then it follows that any other basis is countable as well. In particular. that is. By continuity of the scalar product it follows that M ⊥ is a closed linear subspace and by linearity that . In particular. If fact. if there is one countable basis. any linearly ordered chain has an upper bound (the union of all sets in the chain). the proof requires Zorn’s lemma: The collection of all orthonormal sets in H can be partially ordered by inclusion. it can be shown that L2 (M. Any separable inﬁnite dimensional Hilbert space is unitarily equivalent to 2 (N). Moreover.29) (which are normalized such that Pn (1) = 1). ∀ϕ ∈ M } is called the orthogonal complement of M . The projection theorem and the Riesz lemma Let M ⊆ H be a subset. The resulting polynomials are up to a normalization equal to the Legendre polynomials P0 (x) = 1. (1.5. We will assume all Hilbert spaces to be separable. 1. P2 (x) = 3 x2 − 1 . an orthonormal basis. this set is countable. we infer that j=0 {ϕj }N is an orthonormal basis. Then the map U : H → 2 (N). Let me remark that if H is not separable. j=0 Example. Hence Zorn’s lemma implies the existence of a maximal element.3. Now let {φk }k∈K be a second basis and consider the set Kj = {k ∈ K φk . Moreover. P1 (x) = x. ψ )j∈N is unitary (by Theorem 1. there is only one (separable) inﬁnite dimensional Hilbert space: Let H be an inﬁnite dimensional Hilbert space and let {ϕj }j∈N be any orthogonal basis. there still exists an orthonormal basis.4. ψ = 0.3 (iii)). Hence the set K = j∈J Kj is ˜ ˜ countable as well. Proof. Theorem 1.. If H is separable. But k ∈ K\K implies φk = 0 and hence K = K.. dµ) is separable. In L2 (−1.36 1. up to unitary equivalence. Since {ψj }N is total. ψ → ( ϕj . Hilbert spaces for any ﬁnite n and thus also for N . then M ⊥ = {ψ ϕ. ϕj = 0}. 2 . We know that there is at least one countable orthonormal basis {ϕj }j∈J . Since these are the expansion coeﬃcients of ϕj with ˜ respect to {φk }k∈K . 1) we can orthogonalize the polynomial fn (x) = xn . Theorem 1. then every orthonormal basis is countable. However.
(1. Since M is closed. In turns out that in a Hilbert space every bounded linear functional can be written in this way. that is.2. Then there is a vector ϕ ∈ H such that (ψ) = ϕ. In other words. to every ψ ∈ H we can assign a unique vector ψ which is the vector in M closest to ψ. Note that we have 2 PM = PM (1.3. For every ˜ ψ ∈ H we have (ψ)ϕ − (ϕ)ψ ∈ Ker( ) and hence ˜ ˜ 0 = ϕ. then every ψ ∈ H can be uniquely written as ψ = ψ + ψ⊥ with ψ ∈ M and ψ⊥ ∈ M ⊥ . we can choose ϕ = (ϕ)∗ ϕ. Let M be a closed linear subspace of a Hilbert space H. The rest ψ − ψ lies in M ⊥ . One writes M ⊕ M⊥ = H in this situation. M ⊥⊥ = M . By the CauchySchwarz inequality we know that ϕ : ψ → ϕ. ˜ ˜ The following easy consequence is left as an exercise. it is a Hilbert space and has an orthonormal basis {ϕj }j∈J .32) Note that by H⊥ = {} we see that M ⊥ = {0} if and only if M is dense. Finally we turn to linear functionals.30) and PM ψ. If M is an arbitrary subset we have at least M ⊥⊥ = span(M ). ϕ PM ψ = ψ ⊥ . ψ for all ψ ∈ H. to operators : H → C. Otherwise Ker( ) = {ψ (ψ) = 0} is a proper subspace and we can ﬁnd a unit vector ϕ ∈ Ker( )⊥ . PM ϕ . Proof. a Hilbert space is equivalent to its own dual space H∗ = H. Hence the result follows from Theorem 1.31) since PM ψ. ˜ ˜ ˜ ˜ ˜ In other words. The operator PM ψ = ψ is called the orthogonal projection corresponding to M . For example we have H⊥ = {0} since any vector in H⊥ must be in particular orthogonal to all vectors in some orthonormal basis. ψ . = ψ. ψ is a bounded linear functional (with norm ϕ ). we see that the vectors in a closed subspace M are precisely those which are orthogonal to all vectors in M ⊥ . (ψ)ϕ − (ϕ)ψ = (ψ) − (ϕ) ϕ. In other words. PM ϕ (1. Suppose is a bounded linear functional on a Hilbert space H. Clearly we have PM ⊥ ψ = ψ − Moreover. The projection theorem and the Riesz lemma 37 (span(M ))⊥ = M ⊥ .7 (Riesz lemma). Theorem 1. If ≡ 0 we can choose ϕ = 0. ϕ = ψ.1. ϕ = ψ . (1.6 (projection theorem). Proof. Theorem 1. that is.33) .
Show that an orthogonal projection PM = 0 has norm one. Aϕk with respect to this basis. (1.38 1.6. The elements should be products ψ ⊗ ψ of elements ψ ∈ H .34) Then there is a unique bounded operator A such that B(ψ.35) Problem 1. Problem 1. j=1 ψj 2 j < ∞}.7. Show that P1 ≤ P2 (that is ψ. ψ2 ) ∈ H1 × H2 together with the scalar product (ϕ1 . Show that L(H) is not separable H is inﬁnite dimensional. Problem 1.37) which becomes a Hilbert space with the scalar product ∞ ∞ ∞ ϕj . Hilbert spaces Corollary 1. (1. Compute its range and kernel.2.8. Suppose P1 and P1 are orthogonal projections. Similarly for H2 . Problem 1. ϕ2 ). H1 can be identiﬁed with {(ψ1 . Let {ϕj } be some orthonormal basis. P1 ψ ≤ ψ.38) ˜ Suppose H and H are two Hilbert spaces. ϕ) = Aψ. B(ψ. Our goal is to construct their ˜ tensor product. More generally. (1. ψ1 1 + ϕ2 .8. (ψ1 . (1. that is. j=1 j=1 ψj = j=1 ϕj . Prove Corollary 1. ψ 2 2 . be a countable collection of Hilbert spaces and deﬁne ∞ ∞ ∞ Hj = { j=1 j=1 ψ j  ψ j ∈ Hj . Problem 1.4. Suppose B is a bounded skew liner form. Show that a bounded linear operator A is uniquely determined by its matrix elements Ajk = ϕj . Orthogonal sums and tensor products Given two Hilbert spaces H1 and H2 we deﬁne their orthogonal sum H1 ⊕H2 to be the set of all pairs (ψ1 .36) It is left as an exercise to verify that H1 ⊕ H2 is again a Hilbert space. ψ2 ). ϕ . ψ j j . f (x) → projection. P2 ψ ) is equivalent to Ran(P1 ) ⊆ Ran(P2 ). let Hj j ∈ N.4. Moreover.5. 0)ψ1 ∈ H1 } and we can regard H1 as a subspace of H1 ⊕ H2 .3. 1 2 (f (x) + f (−x)) is a 1. It is also custom to write ψ1 + ψ2 instead of (ψ1 . ψ2 ) = ϕ1 . Problem 1. Show P : L2 (R) → L2 (R). ϕ) ≤ C ψ ϕ . (1.
˜ called the tensor product H ⊗ H ˜ Lemma 1. respectively. since span{ϕj }. it is easy to ˜ ˜ ˜ But the latter is dense in see that ϕj ⊗ ϕk is dense in F(H. respectively. That ϕj ⊗ ϕk is an orthonormal set is immediate from (1. H). ψ⊗(ψ1 + ψ2 ) = ψ⊗ ψ1 +ψ⊗ ψ2 .39) ˜ ˜ ˜ ˜ ˜ ˜ ˜ Since we want (ψ1 +ψ2 )⊗ ψ = ψ1 ⊗ ψ+ψ2 ⊗ ψ. H). Hence we start with the set of all ﬁnite linear combinations of ˜ elements of H × H n ˜ F(H. ϕk are orthonormal bases for H. ψ = j. d˜ ) = L2 (M × M .k αjk 2 > 0. d˜ ) ⊆ L2 (M × M . More˜ ˜ over. H). dµ) ⊗ L2 (M . span{ϕk } is dense in H. Orthogonal sums and tensor products 39 ˜ ˜ and ψ ∈ H. ψi ˜ ˜ (1.k=1 ˜ αj βk (ψj . d˜) be two measure spaces. ˜ obtain a scalar product.40) ˜ ˜ and write ψ ⊗ ψ for the equivalence class of (ψ. ˜ µ Example. k=1 ˜ βk ψk )} (1. Example.9. we need to ensure positivity. We have H ⊗ Cn = Hn . 0 and pick orthonormal bases ϕj . Then we have 2 (M. dµ) and (M . where ˜ ˜ ˜ and (αψ) ⊗ ψ n n n ˜ N (H. dµ) ⊗ L2 (M . αjk = ˜ αi ϕj . ˜ Proof. (1. If ϕj . H. αj ∈ C}. Let (M. ψj )(ψj .44) . ˜ = ψ ⊗ (αψ) we consider F(H. dµ) ⊗ L2 (M .42) j. φ (1. H. Now ˜ µ ˜ Clearly we have L µ 2 (M. Let ψ = i αi ψi ⊗ψi = ˜i }. ψ). ˜ µ ˜ L µ 2 (M. Next we deﬁne ˜ ˜ ˜ ˜ ψ ⊗ ψ. Then M ˜ M (ϕj (x)ϕk (y))∗ f (x. H) with respect to the induced norm is ˜ of H and H. φ ψ. H)/N (H. H)/N (H. dµ × d˜ ). H) = { j=1 ˜ ˜ ˜ αj (ψj . ϕk for span{ψi }. dµ × d˜ ). ψi ϕk . respectively. ψj ) ∈ H × H.41). H) = span{ j.k i and we compute ψ. span{ψ ˜ Then ψ= αjk ϕj ⊗ ϕk . y)dµ(x)d˜(y) = 0 ˜ µ (1.41) ˜ ˜ To show that we which extends to a skew linear form on F(H. d˜ ) as in our ˜ µ take an orthonormal basis ϕj ⊗ ϕk for L ˜ previous lemma.43) ˜ ˜ The completion of F(H. then ˜ ˜ ϕj ⊗ ϕk is an orthonormal basis for H ⊗ H.1. H)/N (H.4. ψk ) − ( j=1 αj ψj . φ ⊗ φ = ψ. H)/N (H. (1. ˜ ˜ H ⊗ H.
(i) and (ii) are obvious. If H = Cn and A = (ajk )1≤j. Example. Show (1. ψ (compare Corollary 1. (ii) A∗∗ = A.k≤n .k≤n .46) 1. A∗ ψ = Aϕ. (1.e.47) ˜ ˜ Problem 1.e.46) where equality has to be understood in the sense. (iv) A = A∗ and A 2 (1. (iv) follows from A∗ = sup ϕ = ψ =1  ψ. Hence ϕj ⊗ ϕk is a basis for L2 (M × M . then (i) (A + B)∗ = A∗ + B ∗ .49) . y and thus f = 0. We even note ∞ ∞ ( j=1 Hj ) ⊗ H = j=1 (Hj ⊗ H). Let A. But this implies f (x. (AB)ψ = A∗ ϕ. It is straightforward to extend the tensor product to any ﬁnite number of Hilbert spaces.8). x and ˜ µa. ψ .40 1. kj Lemma 1. The C ∗ algebra of bounded linear operators We start by introducing a conjugation for operators on a Hilbert space H.e. (iii) (AB)∗ = B ∗ A∗ . M fk (x) = ˜ M ϕk (y)∗ f (x. x. ϕ  = A . that both spaces are unitarily equivalent by virtue of the identiﬁcation ∞ ∞ ( j=1 ψj ) ⊗ ψ = j=1 ψj ⊗ ψ.10. y) = 0 for µa. Hilbert spaces implies ϕj (x)∗ fk (x)dµ(x) = 0. = A∗ A = AA∗ . Proof. then the adjoint operator is deﬁned via ϕ. Bψ = B ∗ A∗ ϕ.45) and hence fk (x) = 0 µa. (1. ˜ such that ψ = αφ and ψ Problem 1.48) (αA)∗ = α∗ A∗ . dµ × d˜) ˜ ˜ µ and equality follows. (1. then A∗ = (a∗ )1≤j. We have ψ⊗ ψ = φ⊗ φ if and only if there is some α ∈ C\{0} ˜ = α−1 φ.9.8. Let A = L(H). B ∈ L(H). A∗ ϕ  = sup ϕ = ψ =1  Aψ. y)d˜(y) ˜ µ (1.5. (iii) follows from ϕ.
2 (N) ∀ψ ∈ H. Show that A is normal if and only if Aψ = A∗ ψ .11.51) a 2 = a∗ a (1. . Any subalgebra which is also closed under involution. a∗∗ = a. Compute the adjoint of S : (0. selfadjoint if a = a∗ . An ideal is a subspace I ⊆ A such that a ∈ I. (Hint: Problem 0. where we have used ϕ = sup As a consequence of A∗ continuous. Note that if there is and identity e we have e∗ = e and hence (a−1 )∗ = (a∗ )−1 (show this). . is called a ∗algebra. (a1 . If it is closed under the adjoint map it is called a ∗ideal. Weak and strong convergence 41 and A∗ A = = sup ϕ = ψ =1  ϕ. Weak and strong convergence Sometimes a weaker notion of convergence is useful: We say that ψn converges weakly to ψ and write wlim ψn = ψ n→∞ or ψn ψ. a1 . An element a ∈ A is called normal if aa∗ = a∗ a. Clearly both selfadjoint and unitary elements are normal.6. a2 .10. ψ. (orthogonal) projection if a = a∗ = a2 . b ∈ A implies ab ∈ I and ba ∈ I. satisfying ψ =1  = A observe that taking the adjoint is In general.1. (ab)∗ = b∗ a∗ . a Banach algebra A together with an involution (αa)∗ = α∗ a∗ . . Problem 1. . The element a∗ is called the adjoint of a.53) → 2 (N). (1.6) Problem 1. a3 . . ).12. (a + b)∗ = a∗ + b∗ . Let A ∈ L(H).6. Problem 1. unitary if aa∗ = a∗ a = I.50) sup ϕ =1 = A 2. Note that a∗ = a follows from (1.54) . The continuous function C(I) together with complex conjugation form a commutative C ∗ algebra. a2 . Show that U : H → H is unitary if and only if U −1 = U ∗ . Aψ  (1. and positive if a = bb∗ for some b ∈ A. Example. (1.52) is called a C ∗ algebra.52) and aa∗ ≤ a a∗ . (1. ) → 1. ϕ . A∗ Aψ  = Aϕ 2 sup ϕ = ψ =1  Aϕ.
Lemma 1. ψn → ϕ. However. (ψ − ψ) = 0. Let ϕn be an (inﬁnite) orthonormal set. A sequence ψn is called weak Cauchy sequence if ϕ.42 1. Clearly an orthonormal basis does not have a norm convergent subsequence. Let H be a Hilbert space. (i) Observe ψ 2 ψ we have: ψn → ψ if = ψ.56) The converse is straightforward. since ϕn = 1. ψ implies ϕ. (iv) For a weakly convergent sequence ψn and only if lim sup ψn ≤ ψ .) Clearly ψn → ψ implies ψn ψ and hence this notion of convergence is indeed weaker. Hilbert spaces if ϕ. (ϕn does not converge to 0. ψn ) + ψn 2 → 0. ψn → ϕ. since ϕ. ψn ≤ ψ lim inf ψn . Proof. ψn → ϕ. (1.11. ψn . Hence the unit ball in an inﬁnite dimensional Hilbert space is never compact. let me remark that similar concepts can be introduced for operators. (iv) By (i) we have lim ψn = ψ and hence ψ − ψn 2 = ψ 2 − 2Re( ψ.12. we can at least extract weakly convergent subsequences: Lemma 1. ψ for every ϕ ∈ H (show that a weak limit is unique). Finally. where convergence in the operator norm makes no sense at all. .55) (ii) For every ϕ we have that  ϕ. ψnm converges for all k. ϕn → 0 for every ψ since these are just the expansion coeﬃcients of ψ. ϕ. Every bounded sequence ψn has weakly convergent subsequence. the weak limit is unique. Then ψ = m cm ϕm is the desired limit. Moreover. This is of particular importance for the case of unbounded operators. ψn is Cauchy for every ϕ ∈ H. Example. . (1. ψ ˜ ˜ and ϕ. ψ = lim inf ψ. Then ψ. ψnm converges for every ϕ ∈ H and hence ψnm is a weak Cauchy sequence. (i) ψn ψ implies ψ ≤ lim inf ψn . then by the usual diagonal sequence argument we can ﬁnd a subsequence ψnm such that ϕk . ψn  ≤ C(ϕ) is bounded. Since ψn is bounded. (iii) Every weak Cauchy sequence converges weakly. (iii) Let ϕm be an orthonormal basis and deﬁne cm = limn→∞ ϕm . Hence by the uniform boundedness principle we have ψn = ψn . Let H be a Hilbert space. Proof. ≤ C. Let ϕk be an ONB. (ii) Every weak Cauchy sequence ψn is bounded: ψn ≤ C.
Suppose An is a sequence of bounded operators. . (ii) Just use An ψ − Aψ ≤ An ψ − An ϕ + An ϕ − Aϕ + Aϕ − Aψ (1. . wlim An = A n→∞ :⇔ An ψ Aψ ∀ψ ∈ D(A) ⊆ D(An ). (1.58) Clearly norm convergence implies strong convergence and strong convergence implies weak convergence.11 (i).1. 0.57) It is said to converge weakly to A. (1. x2 . slim An = A n→∞ :⇔ An ψ → Aψ ∀x ∈ D(A) ⊆ D(An ). .62) and hence A∗ in this case. ) = (xn+1 . . . x2 . The same result holds if strong convergence is replaced by weak convergence. A∗ ψ = An ϕ. (1.59) ∗ and the operator Sn ∈ L( 2 (N)) which shifts a sequence n places to the right and ﬁlls up the ﬁrst n places with zeros ∗ Sn (x1 . . . . ). x1 . A∗ ψ n and hence A∗ n s A∗ → n (1. Consider the operator Sn ∈ L( 2 (N)) which shifts a sequence n places to the left Sn (x1 . Thus at least for normal operators taking adjoints is continuous with respect to strong convergence. than slimn→∞ An = A. . xn+2 . . Proof. (1.63) ≤ 2C ψ − ϕ + An ϕ − Aϕ . if An and A are normal we have (An − A)∗ ψ = (An − A)ψ (1. x2 . ψ = ϕ. (iii) If An y → Ay for y in a dense set and An ≤ C. . However. ψ → Aϕ.13. . ). .60) Note that this example also shows that taking adjoints is not continuous s with respect to strong convergence! If An → A we only have ϕ.6. (i) and (ii) follow as in Lemma 1. . Example. (ii) Every strong Cauchy sequence An is bounded: An ≤ C. Lemma 1. ) = (0. . Weak and strong convergence 43 A sequence of operators An is said to converge strongly to A. n places ∗ Then Sn converges to zero strongly but not in norm and Sn converges weakly to zero but not strongly. .61) A∗ in general. (i) slimn→∞ An = A implies A ≤ lim inf An .
Let {ϕj } be some orthonormal basis.14 (Stone–Weierstraß. A subspace M ⊆ H is closed if and only if every weak Cauchy sequence in M has a limit in M . we also have max{f. If F ⊂ C(K.14. Show that ψn ϕ if and only if ψn is bounded and ϕj . Let {ϕj }∞ be some orthonormal basis and deﬁne j=1 ∞ ψ Show that . Then ψn . Suppose K is a compact set and let C(K. 2 min{f. the spectral theorem will show that the closed ∗algebra generated by this operator is isomorphic to the C ∗ algebra of continuous functions C(K) over some compact set. ϕn → ψ.13) 1 there is a polynomial pn (t) such that t−pn (t) < n and hence pn (f ) → f . if f. ψn → ϕj .z (y) = f (y) and fy. Problem 1. for every x1 = x2 there is some function f ∈ F such that f (x1 ) = f (x2 )). We need to ﬁnd some fε ∈ A with f − fε ∞ (f + g) + f − g . R).7. then the algebra generated by F is dense. g} = (f + g) − f − g 2 (1. 2j ϕ if and only if ψn − ψ (1. Show that ψn → 0. since A separates points. g in A. g} = in A. First of all. Denote by A the algebra generated by F . Problem 1. R) be the Banach algebra of continuous functions (with the sup norm). observe that for given y.. The case of weak convergence is left as an exercise.13.e. ϕ . Proof.16. z ∈ K there is a function fy. Hence it is important to be able to identify dense sets: Theorem 1. Hilbert spaces and choose ϕ in the dense subspace such that ψ − ϕ ≤ ε such that An ϕ − Aϕ ≤ 2 .64) w is a norm. Now ﬁx f ∈ C(K.44 1. Problem 1. real version). In particular.15.z ∈ A such that fy. Suppose ψn → ψ and ϕn ε 4C and n large ϕ. Show that this is wrong without the boundedness assumption.z (z) = f (z) . w w = j=1 1  ϕj . Note that if f ∈ A. Problem 1. (Hint: M = M ⊥⊥ .65) < ε.) 1. ψ for every j. we have f  ∈ A: By the Weierstraß approximation theorem (Theorem 0. R) contains the identity 1 and separates points (i. Appendix: The Stone–Weierstraß theorem In case of a selfadjoint operator. ψ .
.7. Suppose K is a compact set and let C(K) be the C ∗ algebra of continuous functions (with the sup norm). Then fz = max{fy1 .68) (1. . since pn can be chosen to contain no constant term). If there is such a t0 . cover K. If F ⊂ C(K) separates points. There are two possibilities. say U (y1 ). but they are not dense (since the uniform limit of holomorphic functions is again holomorphic). . . Appendix: The Stone–Weierstraß theorem 45 (show this). the identity is clearly missing from A. . Proof.16. Just observe that F = {Re(f ). .66) (1. . Hence any realvalued continuous functions can be ˜ approximated by elements from F . . then the closure of the ∗algebra generated by F is either C(K) or {f ∈ C(K)f (t0 ) = 0} for some t0 ∈ K. . . . Theorem 1. for every y ∈ K there is a neighborhood U (y) such that fy. If there is no such point we can proceed as in the proof of the Stone–Weierstraß theorem to show that the identity can be approximated by elements in A (note that to show f  ∈ A if f ∈ A we do not need the identity. Note that the additional requirement of being closed under complex conjugation is crucial: The functions holomorphic on the unit ball and continuous on the boundary separate points. ˜ Proof. either all f ∈ F vanish at one point t0 ∈ K (there can be at most one such point since F separates points) or there is no such point. fzk } ∈ A satisﬁes fε < f + ε. . Im(f )f ∈ F } satisﬁes the assumption of the real version.69) and a corresponding ﬁnite cover V (z1 ). Corollary 1. Now fε = min{fz1 . x ∈ V (z) (1. in particular this holds for the real and imaginary part for any given complexvalued function. then the ∗algebra generated by F is dense.1. However. Since f − ε < fzl < fε . x ∈ U (y) (1.15 (Stone–Weierstraß). . we have found a required function.z } ∈ A and satisﬁes fz > f − ε by construction. Next.67) and since K is compact. U (yj ). V (zk ). If F ⊂ C(K) contains the identity 1 and separates points. fyj . ﬁnitely many.z (x) > f (x) − ε. adding the identity to A we get A + C = C(K) and it is easy to see that A = {f ∈ C(K)f (t0 ) = 0}. . Suppose K is a compact set and let C(K) be the C ∗ algebra of continuous functions (with the sup norm). . Since fz (z) = f (z) for every z ∈ K there is a neighborhood V (z) such that fz (x) < f (x) + ε.z .
46
1. Hilbert spaces
Problem 1.17. Show that the of functions ϕn (x) = an orthonormal basis for H = L2 (0, 2π).
√1 ein x , 2π
n ∈ Z, form
1 Problem 1.18. Show that the ∗algebra generated by fz (t) = t−z , z ∈ C, is dense in the C ∗ algebra C∞ (R) of continuous functions vanishing at inﬁnity. (Hint: Add ∞ to R to make it compact.)
Chapter 2
Selfadjointness and spectrum
2.1. Some quantum mechanics
In quantum mechanics, a single particle living in R3 is described by a complexvalued function (the wave function) ψ(x, t), (x, t) ∈ R3 × R, (2.1)
where x corresponds to a point in space and t corresponds to time. The quantity ρt (x) = ψ(x, t)2 is interpreted as the probability density of the particle at the time t. In particular, ψ must be normalized according to ψ(x, t)2 d3 x = 1,
R3
t ∈ R.
(2.2)
The location x of the particle is a quantity which can be observed (i.e., measured) and is hence called observable. Due to our probabilistic interpretation it is also a random variable whose expectation is given by Eψ (x) = xψ(x, t)2 d3 x.
R3
(2.3)
In a real life setting, it will not be possible to measure x directly and one will only be able to measure certain functions of x. For example, it is possible to check whether the particle is inside a certain area Ω of space (e.g., inside a detector). The corresponding observable is the characteristic function χΩ (x) of this set. In particular, the number Eψ (χΩ ) = χΩ (x)ψ(x, t)2 d3 x =
R3 Ω
ψ(x, t)2 d3 x
(2.4) 47
48
2. Selfadjointness and spectrum
corresponds to the probability of ﬁnding the particle inside Ω ⊆ R3 . An important point to observe is that, in contradistinction to classical mechanics, the particle is no longer localized at a certain point. In particular, the meansquare deviation (or variance) ∆ψ (x)2 = Eψ (x2 ) − Eψ (x)2 is always nonzero. In general, the conﬁguration space (or phase space) of a quantum system is a (complex) Hilbert space H and the possible states of this system are represented by the elements ψ having norm one, ψ = 1. An observable a corresponds to a linear operator A in this Hilbert space and its expectation, if the system is in the state ψ, is given by the real number Eψ (A) = ψ, Aψ = Aψ, ψ , (2.5) where ., .. denotes the scalar product of H. Similarly, the meansquare deviation is given by ∆ψ (A)2 = Eψ (A2 ) − Eψ (A)2 = (A − Eψ (A))ψ 2 . (2.6)
Note that ∆ψ (A) vanishes if and only if ψ is an eigenstate corresponding to the eigenvalue Eψ (A), that is, Aψ = Eψ (A)ψ. From a physical point of view, (2.5) should make sense for any ψ ∈ H. However, this is not in the cards as our simple example of one particle already shows. In fact, the reader is invited to ﬁnd a square integrable function ψ(x) for which xψ(x) is no longer square integrable. The deeper reason behind this nuisance is that Eψ (x) can attain arbitrary large values if the particle is not conﬁned to a ﬁnite domain, which renders the corresponding operator unbounded. But unbounded operators cannot be deﬁned on the entire Hilbert space in a natural way by the closed graph theorem (Theorem 2.7 below) . Hence, A will only be deﬁned on a subset D(A) ⊆ H called the domain of A. Since we want A to at least be deﬁned for most states, we require D(A) to be dense. However, it should be noted that there is no general prescription how to ﬁnd the operator corresponding to a given observable. Now let us turn to the time evolution of such a quantum mechanical system. Given an initial state ψ(0) of the system, there should be a unique ψ(t) representing the state of the system at time t ∈ R. We will write ψ(t) = U (t)ψ(0). (2.7)
Moreover, it follows from physical experiments, that superposition of states holds, that is, U (t)(α1 ψ1 (0) + α2 ψ2 (0)) = α1 ψ1 (t) + α2 ψ2 (t) (α1 2 + α2 2 = 1). In other words, U (t) should be a linear operator. Moreover,
2.1. Some quantum mechanics
49
since ψ(t) is a state (i.e., ψ(t) = 1), we have U (t)ψ = ψ . (2.8)
Such operators are called unitary. Next, since we have assumed uniqueness of solutions to the initial value problem, we must have U (0) = I, U (t + s) = U (t)U (s). (2.9)
A family of unitary operators U (t) having this property is called a oneparameter unitary group. In addition, it is natural to assume that this group is strongly continuous
t→t0
lim U (t)ψ = U (t0 )ψ,
ψ ∈ H.
(2.10)
Each such group has an inﬁnitesimal generator deﬁned by 1 D(H) = {ψ ∈ H lim (U (t)ψ − ψ) exists}. t→0 t (2.11) This operator is called the Hamiltonian and corresponds to the energy of the system. If ψ(0) ∈ D(H), then ψ(t) is a solution of the Schr¨dinger o equation (in suitable units) i d ψ(t) = Hψ(t). dt (2.12) i Hψ = lim (U (t)ψ − ψ), t→0 t
This equation will be the main subject of our course. In summary, we have the following axioms of quantum mechanics. Axiom 1. The conﬁguration space of a quantum system is a complex separable Hilbert space H and the possible states of this system are represented by the elements of H which have norm one. Axiom 2. Each observable a corresponds to a linear operator A deﬁned maximally on a dense subset D(A). Moreover, the operator corresponding to a polynomial Pn (a) = n αj aj , αj ∈ R, is Pn (A) = n αj Aj , j=0 j=0 D(Pn (A)) = D(An ) = {ψ ∈ D(A)Aψ ∈ D(An−1 )} (A0 = I). Axiom 3. The expectation value for a measurement of a, when the system is in the state ψ ∈ D(A), is given by (2.5), which must be real for all ψ ∈ D(A). Axiom 4. The time evolution is given by a strongly continuous oneparameter unitary group U (t). The generator of this group corresponds to the energy of the system. In the following sections we will try to draw some mathematical consequences from these assumptions:
50
2. Selfadjointness and spectrum
First we will see that Axiom 2 and 3 imply that observables correspond to selfadjoint operators. Hence these operators play a central role in quantum mechanics and we will derive some of their basic properties. Another crucial role is played by the set of all possible expectation values for the measurement of a, which is connected with the spectrum σ(A) of the corresponding operator A. The problem of deﬁning functions of an observable will lead us to the spectral theorem (in the next chapter), which generalizes the diagonalization of symmetric matrices. Axiom 4 will be the topic of Chapter 5.
2.2. Selfadjoint operators
Let H be a (complex separable) Hilbert space. A linear operator is a linear mapping A : D(A) → H, (2.13) where D(A) is a linear subspace of H, called the domain of A. It is called bounded if the operator norm A = sup
ψ =1
Aψ =
sup
ϕ = ψ =1
 ψ, Aϕ 
(2.14)
is ﬁnite. The second equality follows since equality in  ψ, Aϕ  ≤ ψ Aϕ is attained when Aϕ = zψ for some z ∈ C. If A is bounded it is no restriction to assume D(A) = H and we will always do so. The Banach space of all bounded linear operators is denoted by L(H). The expression ψ, Aψ encountered in the previous section is called the quadratic form qA (ψ) = ψ, Aψ , ψ ∈ D(A), (2.15)
associated to A. An operator can be reconstructed from its quadratic form via the polarization identity ϕ, Aψ = 1 (qA (ϕ + ψ) − qA (ϕ − ψ) + iqA (ϕ − iψ) − iqA (ϕ + iψ)) . (2.16) 4
A densely deﬁned linear operator A is called symmetric (or Hermitian) if ϕ, Aψ = Aϕ, ψ , ψ, ϕ ∈ D(A). (2.17) The justiﬁcation for this deﬁnition is provided by the following Lemma 2.1. A densely deﬁned operator A is symmetric if and only if the corresponding quadratic form is realvalued.
(Multiplication operator) Consider the multiplication operator D(A) = {f ∈ L2 (Rn . ψ = Im ϕ. ˜ By Axiom 2. In other words. Clearly (2. (2. In addition. (2.2) Ker(A∗ ) = Ran(A)⊥ . (2. Selfadjoint operators 51 Proof. (Af )(x) = A(x)f (x).21) For symmetric operators we clearly have A ⊆ A∗ .19) This already narrows the class of admissible operators to the class of symmetric operators by Axiom 3. we write A = A ˜ ˜ D(A) ⊆ D(A) ˜ ˜ if both A ⊆ A and A ⊆ A hold. Aψ ) (2.1). ψ ∈ D(A). dµ)}. Aϕ − ϕ. Replacing ϕ by iϕ in this last equation shows Im Aϕ. consider Ωn = {x ∈ Rn  A(x) ≤ n} Rn . ψ . note that (Problem 2. Here we write A ⊆ A if ˜ and Aψ = Aψ for all ψ ∈ D(A). which is a special case of a multiplication operator. . First of all note that D(A) is dense. equality will not hold in general unless one operator is bounded (Problem 2. ∀ϕ ∈ D(A)} . Clearly we have (αA)∗ = α∗ A∗ for α ∈ C and (A + B)∗ ⊇ A∗ + B ∗ provided D(A + B) = D(A) ∩ D(B) is dense. Example. if A is another ˜ ˜ ˜ symmetric operator such that A ⊆ A. ϕ . taking the imaginary part of the identity qA (ψ + iϕ) = qA (ψ) + qA (ϕ) + i( ψ. ψ = Re ϕ.22) given by multiplication with the measurable function A : Rn → C. for every f ∈ L2 (Rn . Conversely. In fact. then A = A. However. Aψ and ﬁnishes the proof. The adjoint operator A∗ of a densely deﬁned linear operator A is deﬁned by ˜ ˜ D(A∗ ) = {ψ ∈ H∃ψ ∈ H : ψ.18) shows Re Aϕ. dµ) the function fn = χΩn f ∈ D(A) converges to f as n → ∞ by dominated convergence. This is for example true in the case of the position operator x. However. Then. (2. a densely deﬁned operator A is symmetric if and only if ψ.17) implies that Im(qA (ψ)) = 0. For later use. Aψ = Aψ. let us tackle the issue of the correct domain. Aψ . If in addition.2. dµ)  Af ∈ L2 (Rn .20) ˜ A∗ ψ = ψ The requirement that D(A) is dense implies that A∗ is welldeﬁned. Our goal is to show that observables correspond to selfadjoint operators. note that D(A∗ ) might not be dense in general. Next. A = A∗ holds. then A is called selfadjoint.2. A should be deﬁned maximally. it might contain no vectors other than 0. Aϕ = ψ. In fact. that is.
Furthermore. our calculation only shows A ⊆ A∗ . f ∈ D(A) that h. But for our calculation we had to assume h ∈ D(A) and there might be some functions in D(A∗ ) which do not satisfy this requirement! In ˜ particular.20) it is clear that A ⊆ A∗∗ and . we need to work a little harder: If h ∈ D(A∗ ) there is some g ∈ L2 (Rn .28) that is. increasing the domain of A implies decreasing the domain of A∗ . A is selfadjoint if A is realvalued. we infer A ⊆ B ⊆ B ∗ ⊆ A∗ = A implying A = B. (2. that is. we even have h(x)A(x)∗ = g(x) ∈ L2 (Rn . (2. Thus there is no point in trying to extend the domain of a selfadjoint operator further. ˜ ˜ ˜ (Af )(x) = A(x)∗ f (x). D(A) = {f ∈ L2 (Rn . Selfadjointness and spectrum Next. Such operators are called normal. From the deﬁnition (2. dµ) such that h(x)∗ A(x)f (x)dµ(x) = and thus (h(x)A(x)∗ − g(x))∗ f (x)dµ(x) = 0.23) ˜ where A is multiplication by A(x)∗ . if A∗ is densely deﬁned (which is the case if A is symmetric) we can consider A∗∗ . In fact. if A is selfadjoint and B is a symmetric extension. dµ)  Af ∈ L2 (Rn .27) f ∈ D(A). At ﬁrst sight this seems to show that the adjoint of Note D(A) ˜ A is A. f .2. f ∈ L2 (Rn . To show that equality holds. dµ) vanishes. (2. dµ) and thus A∗ is multiplication by A(x)∗ and D(A∗ ) = D(A). In particular. (2.52 2. (2. f ∈ D(A). In particular. Now note that A ⊆ B ⇒ B ∗ ⊆ A∗ .26) g(x)∗ f (x)dµ(x). Performing a formal computation we have for h. Af = h(x)∗ A(x)f (x)dµ(x) = ˜ (A(x)∗ h(x))∗ f (x)dµ(x) = Ah. In the general case we have at least Af = A∗ f for all f ∈ D(A) = D(A∗ ). dµ)}.24) ˜ = D(A). Corollary 2.25) which shows that χΩn (h(x)A(x)∗ − g(x))∗ ∈ L2 (Rn . Since n is arbitrary. Selfadjoint operators are maximal. (2. χΩn (x)(h(x)A(x)∗ − g(x))∗ f (x)dµ(x) = 0. dµ). they do not have any symmetric extensions. let us compute the adjoint of A.
6). that is. Integration by parts shows ˜ 2π x ∗ f (x) g(x) − i 0 x 0 g (t)dt ˜ dx = 0 (2.2. 2π). In summary. dx D(A0 ) = {f ∈ C 1 [0. ϕ ∈ D(A∗ ) implies that A is symmetric since A∗ ϕ = Aϕ for ϕ ∈ D(A).30) holds with g = −ig and we conclude ˜ A∗ f = −i 0 d f. 2π]  f (0) = f (2π) = 0}. ˜ 0 0 for every g ∈ H 1 (0. ϕ for all ψ ∈ D(A). (2.30) for some g ∈ L2 (0. 2π]. b] = {f ∈ C[a. 2π]f ∈ L2 (0. 2π) 0 h(t)dt = 0} implying g(x) = g(0) + i 0 g (t)dt since ⊥ and {1}⊥⊥ = span{1}.33) In particular.29) That A0 is symmetric can be shown by a simple integration by parts (do this). where x AC[a. g ∈ D(A∗ ) implies g ∈ AC[0. A0 f − A∗ g. If g ∈ D(A∗ ) we 0 0 must have 2π 0 g(x)∗ (−if (x))dx = 0 2π g (x)∗ f (x)dx ˜ (2.31) ˜ and hence g(x) − i 0 g (t)dt ∈ {f f ∈ D(A0 )}⊥ . h = 0} = {1} g ∈ AC[0. Moreover. dx D(A∗ ) = H 1 (0. 2π). Since A∗∗ ⊆ A∗ we compute 0 = g.34) . This extension is closely related to extending a linear subspace M via M ⊥⊥ = M (as we will see a bit later) and thus is called the closure A = A∗∗ of A. 2π] and A∗ g = g = −ig . A∗ ϕ = Aψ. A is symmetric but not selfadjoint. b]f (x) = f (a) + a g(t)dt. Thus {f f ∈ D(A0 )} = {h ∈ H 1. (i). Note that the boundary conditions f (0) = f (2π) = 0 are chosen such that the boundary terms occurring from integration by parts vanish.2. 2π)} (2. 2π) = {f ∈ AC[0. But {f f ∈ D(A0 )} = x 2π ˜ {h ∈ C(0. A lies between A and A∗ . If A is symmetric we have A ⊆ A∗ and hence A = A∗∗ ⊆ A∗ . g ∈ L1 (a. Example. Conversely.32) denotes the set of all absolutely continuous functions (see Section 2. However. ψ. (Diﬀerential operator) Take H = L2 (0. this will also follow once we have computed A∗ . Selfadjoint operators 53 thus A∗∗ is an extension of A. f = i(f (0)g(0)∗ − f (2π)g(2π)∗ ) 0 (2. 0 (2. Consider the operator A0 f = −i d f. b)} (2. 2π).
the ﬁrst operator A0 has no eigenvectors at all (i. Selfadjointness and spectrum and since the boundary values of g ∈ D(A∗ ) can be prescribed arbitrary. Now let us take d f. Af − A∗ g. The general solution of the diﬀerential equation is u(x) = u(0)eizx and the boundary conditions imply u(x) = 0.37) Similarly. dx D(A∗ ) = {f ∈ H 1 (0. D(A) = {f ∈ C 1 [0. Thus there are two possibilities. Thus A∗ ⊆ A∗ and we compute 0 Af = −i 0 = g. Thus A0 f = −i d f. f = if (0)(g(0)∗ − g(2π)∗ ).38) (2. the converse is true: For example. 2π)  f (0) = f (2π)}.36) dx which is clearly an extension of A0 . that is the same diﬀerential equation as before. Now we look for solutions of Au = zu. z ∈ C. that is. a solution of the ordinary diﬀerential equation u (x) = zu(x) (2. . By deﬁnition an eigenvector is a (nonzero) solution of A0 u = zu. we see that all eigenvectors are given by 1 n ∈ Z.35) (ii). Hence there are no eigenvectors. 2π]  f (0) = f (2π)}. A = A∗ and thus A is selfadjoint.e. In particular. solutions of the equation A0 ψ = zψ.39) satisfying the boundary conditions u(0) = u(2π) = 0 (since we must have u ∈ D(A0 ). dx D(A0 ) = {f ∈ D(A∗ )  f (0) = f (2π) = 0}. Again the general solution is u(x) = u(0)eizx and the boundary condition requires u(0) = u(0)e2πiz . Either u(0) = 0 (which is of no use for us) or z ∈ Z. but now subject to the boundary condition u(0) = u(2π).54 2. 0 (2..40) un (x) = √ einx . One might suspect that there is no big diﬀerence between the two symmetric operators A0 and A from the previous example. since they coincide on a dense set of vectors. (2. z ∈ C) whereas the second one has an orthonormal basis of eigenvectors! Example. (2. (i). as before. we 0 must have f (0) = f (2π) = 0. However. (2. 2π which are wellknown to form an orthonormal basis. Since this must hold for all f ∈ D(A) we conclude g(0) = g(2π) and A∗ f = −i d f. (ii). Compute the eigenvectors of A0 and A from the previous example.
with the graph norm ψ 2 Γ(A) = ψ closed). This situation is more delicate and will be investigated in Section 2. which is the case if and only if the graph of A is closed. (A + z)ϕ . That is.2. We will use a diﬀerent approach which avoids the use of the adjoint operator. A is called closed if A = A. Two cases need to be distinguished. In this case A is called closable and the ˜ (ψn .2. On the other hand..4. Equivalently.5. For Aψ to be welldeﬁned. Let ψ ∈ D(A∗ ) and A∗ ψ = ψ. ϕ = (A + z ∗ )ϑ. there is a ∗ )ϑ = ψ + z ∗ ψ. if a given symmetric operator A turns out not to be selfadjoint. If A is selfadjoint.3. ϕ ∈ D(A).2). In this case A is called essentially selfadjoint and D(A) is called a core for A. Hence it will be important to know whether a given operator is selfadjoint or not. a criterion for selfadjointness not involving A∗ will be useful. To proceed further. A bounded operator is closed if and only if its domain is closed (show this!). A is closed if and only if Γ(A) equipped 2 + Aψ 2 is a Hilbert space (i. Let A be symmetric such that Ran(A + z) = Ran(A + z ∗ ) = H for one z ∈ C. Since Ran(A + z ∗ ) = H. Clearly.e. We will establish equivalence with our original deﬁnition in Lemma 2. However. The simplest way of extending an operator A is to take the closure of its ˜ graph Γ(A) = {(ψ. Aψn ) → (0. Aψn ) → (ψ. then there is only one selfadjoint extension (if B is another one. Otherwise there might be more than one selfadjoint extension or none at all. ˜ Proof.41) . Then A is selfadjoint. Since we have seen that computing A∗ is not always easy. if (ψn . (A + z)ϕ = ψ + z ∗ ψ. we have A ⊆ B and hence A = B by Corollary 2. ψ) unique operator A which satisﬁes Γ(A) = Γ(A) is called the closure of A. we will need more information on the closure of an operator. Lemma 2. (2. this raises the question of selfadjoint extensions. and hence ψ = ϑ ∈ D(A) since Ran(A + z) = H. but computing the adjoint of an operator is a nontrivial job even in simple situations. Aψ)ψ ∈ D(A)} ⊂ H2 . ψ) ˜ we might try to deﬁne Aψ = ψ. we need that ˜ implies ψ = 0. Now we compute ˜ ϑ ∈ D(A) such that (A + z ˜ ψ. ϕ = ϑ. Selfadjoint operators 55 We will see a bit later that this is a consequence of selfadjointness of A. Our example shows that symmetry is easy to check (in case of diﬀerential operators it usually boils down to integration by parts). we will learn soon that selfadjointness is a much stronger property than symmetry justifying the additional eﬀort needed to prove it.
Let us consider the following two unitary operators from H2 to itself U (ϕ. (2.46) (2. Let us compute the closure of the operator A0 from the previous example without the use of the adjoint operator. ψ) = (ψ.43) also shows A∗ = A∗ . −1 If A is closable and A is injective.56 2. ψ) Γ(A) = 0 ∀(ψ.42) (2. then (A∗ )−1 = (A−1 )∗ .43) and comparing with the last formula shows A∗∗ = A. Selfadjointness and spectrum Example. From Γ(A) = Γ(A)⊥⊥ = (U Γ(A∗ ))⊥ ˜ ˜ = {(ψ. (ii). −ϕ). A∗ ϕ − ψ. and only if D(A Moreover. ϕ) ∈ H  (−ϕ. (i). ϕ). Next note that (provided A is injective) Γ(A−1 ) = V Γ(A). Proof. (iii) If A is injective and the Ran(A) is dense. ψ) ∈ Γ(A)} ˜ ˜ = U (Γ(A)⊥ ) = (U Γ(A))⊥ we conclude that A∗ is closed. (i) A∗ is closed.45) . (ψ. replacing A by A∗ in (2. Lemma 2. ϕ = 0. ψ) ψ. (ii) A is closable if and only if D(A∗ ) is dense and A = A∗∗ respectively (A)∗ = A∗ in this case. Suppose A is a densely deﬁned operator. In this case. ∀ϕ ∈ D(A∗ )} (2. Let f ∈ D(A0 ) and let fn ∈ D(A0 ) be a sequence such that fn → f . Next.4. A0 fn → −ig. 2π] and f (0) = 0.44) ˜ ˜ we see that (0.43) V (ϕ. ϕ) ∈ H2  ϕ. then Ker(A∗ ) = Ran(A)⊥ = {0} and Γ((A∗ )−1 ) = V Γ(A∗ ) = V U Γ(A)⊥ = U V Γ(A)⊥ = U (V Γ(A))⊥ (2. equation (2. Then x fn → g and hence f (x) = 0 g(t)dt. Hence if Ran(A) is dense. then A = A−1 . Hence A is closable if ∗ ) is dense. Aψ = ϕ. Thus f ∈ AC[0. ψ ∀ϕ ∈ D(A∗ )} ˜ ˜ 2 ˜ ˜ = {(ϕ. 2π Moreover f (2π) = limn→0 0 fn (t)dt = 0. ϕ). ψ) = (ψ. From Γ(A∗ ) = {(ϕ. let us collect a few important results. any such f can be approximated by functions in D(A0 ) (show this). (iii). ψ) ∈ Γ(A) if and only if ψ ∈ D(A∗ )⊥ . Conversely.
Replacing z by z ∗ and applying Lemma 2. As noted earlier Ker(A∗ ) = Ran(A)⊥ . (A + ε)ψ 2 ≤ ψ (A + ε)ψ which shows (A + ε)−1 ≤ ε−1 . • Ran(A + z) = Ran(A + z ∗ ) = H.7 (closed graph). 0). Aψ ≥ 0 for all ψ ∈ D(A). Then A∗ is well deﬁned and for all unit vectors . A symmetric operator A is essentially selfadjoint if and only if one of the following conditions holds for one z ∈ C\R. If A is nonnegative. If A is bounded than it is easy to see that Γ(A) is closed. We have A ∈ L(H) if and only if A∗ ∈ L(H).48) we infer that Ker(A−z) = {0} and hence (A−z)−1 exists. • Ker(A∗ + z) = Ker(A∗ + z ∗ ) = {0}. Lemma 2. So let us assume that Γ(A) is closed. if A is closable and A is injective.8 A ∈ L(H).47) If A ∈ L(H) we clearly have D(A∗ ) = H and by Corollary 1. Proof. its domain Ran(A + z) must be equal to H.5. ε > 0. In particular. From (A−z)ψ 2 = (A−x)ψ −iyψ 2 = (A−x)ψ 2 +y 2 ψ 2 ≥ y 2 ψ 2 . In addition. and hence the two conditions are equivalent. By taking the closure of A it is no restriction to assume that A is closed. The argument for the nonnegative case with z < 0 is similar using ε ψ 2 ≤  ψ. setting ψ = (A − z)−1 ϕ (y = 0) shows (A − z)−1 ≤ y−1 .2. we can also admit z ∈ (−∞. Hence (A − z)−1 is bounded and closed. (2. Let z = x + iy. we can also prove the closed graph theorem which shows that an unbounded operator cannot be deﬁned on the entire Hilbert space.2. Let H1 and H2 be two Hilbert spaces and A an operator deﬁned on all of H1 .3 ﬁnishes the general case. Now we can also generalize Lemma 2.3 to the case of essential selfadjoint operators. Moreover. Proof. Theorem 2. Since it is densely deﬁned by assumption. −1 then A = A−1 by Γ(A −1 ) = V Γ(A) = V Γ(A) = Γ(A−1 ). (2. that is ψ.6. Then A is bounded if and only if Γ(A) is closed. Selfadjoint operators 57 shows that (A∗ )−1 = (A−1 )∗ . since A = A∗∗ we obtain Theorem 2. Similarly.
the map (ϕ. That is. positive) if ψ. Aψ ≥ 0 (resp. Let HA be the completion of D(A) with respect to the above scalar product. However. Hence by the uniform boundedness principle there is a constant C such that ∗ ∗ ∗∗ ϕ = A ϕ ≤ C. dµ). This follows from ψn 2 A = ≤ ψn . There is a good way of doing this for nonnegative operators and hence we will consider this case ﬁrst. (2. then it is also Cauchy in H (since ψ ≤ ψ A by assumption) and hence we can identify it with the limit of (ψn ) regarded as a sequence in H. To avoid this. ψ n − ψm ψn A A + ψn . ψ is pointwise bounded (2. we introduce the scalar product ϕ. We claim that HA can be regarded as a subspace of H. (Multiplication operator) Let A be multiplication by A(x) ≥ 0 in L2 (Rn . dµ)} and qA (x) = Rn (2.53) (2. we need to show that if (ψn ) ⊂ D(A) is a Cauchy sequence in HA such that ψn → 0. ψ) → ϕ.49) ϕ (ψ) =  ϕ. For this identiﬁcation to be unique. Example. (A + 1)ψ . Aψ  ≤ Aψ . ψ A − ψ 2 .58 2. Clearly the quadratic form qA can be extended to every ψ ∈ HA by setting qA (ψ) = ψ. n large. (2. Aψ is a scalar product. ψ m A A ψn − ψm + ψn (A + 1)ψm (2.50) deﬁned on D(A). ψ A = ϕ. If A is positive. A ≥ 0. A is bounded and so is A = A . An operator is called nonnegative (resp. ψ ∈ Q(A) = HA . Then Q(A) = D(A1/2 ) = {f ∈ L2 (Rn .54) A(x)f (x)2 dµ(x) . > 0 for ψ = 0) for all ψ ∈ D(A). the diﬃcult part remaining is to show that an observable has at least one selfadjoint extension. dµ)  A1/2 f ∈ L2 (Rn . then ψn A → 0.51) since the right hand side can be made arbitrarily small choosing m. Since selfadjoint operators are already maximal. there might be sequences which are Cauchy with respect to this scalar product but not with respect to our original one. which satisﬁes ψ ≤ ψ A . Finally we want to draw some some further consequences of Axiom 2 and show that observables correspond to selfadjoint operators. If (ψn ) is a Cauchy sequence in D(A). Selfadjointness and spectrum ϕ ∈ D(A∗ ) we have that the linear functional ϕ (ψ) = A∗ ϕ. that is. D(A) ⊆ HA ⊆ H.52) The set Q(A) is also called the form domain of A.
Moreover. Let us deﬁne an operator A by ˜ ˜ D(A) = {ψ ∈ HA ∃ψ ∈ H : ϕ. ˜ (A Now it is time for another Example. π)  f (0) = f (π) = 0}. (2. then there is a non˜ ˜ negative extension A such that Ran(A + 1) = H. ϕ = ψ. ∀ϕ ∈ HA } . Moreover.56) dx2 which corresponds to the onedimensional model of a particle conﬁned to a box. A + 1 is a ˜ onto H.2. Af = − (i). since π g. First of all. ˜ ϕ is bounded linear functional on HA . π) and fn (x) = 0 fn (t)dt . it is straightforward to check that A is symmetric π 0 g(x)∗ (−f )(x)dx = 0 π g (x)∗ f (x)dx = 0 π (−g )(x)∗ f (x)dx.59) we see that fn is Cauchy in HA if and only if both fn and fn are Cauchy x in L2 (0.55) ˜ Since HA is dense. ψ ˜−ψ ˜ Aψ = ψ A ˜ = ϕ. D(A) = {f ∈ C 2 [0.8. Suppose A is a nonnegative operator. (2. π). ˜ ˜ It is also not hard to see that Ran(A + 1) = H. (2. ˜ ˜ + 1)ψ = ψ and hence A + 1 is onto. ϕ A for all ϕ ∈ HA . ˜ Proof.58) (ii).2. ψ is welldeﬁned. ˜ ˜ ψ ∈ HA such that ψ. Indeed. Let us take H = L2 (0. Now we come to our extension result. ψ . Note that A + 1 is injective and ˜ ˜ the best we can hope for is that for a nonnegative extension A. for any ψ ∈ H. By the deﬁnition of A. bijection from D(A) Lemma 2. using integration by parts twice. Thus fn → f and fn → g in L2 (0. (2. (2. Hence there is an element ϕ → ψ. Next let us show HA = {f ∈ H 1 (0. f A = 0 g (x)∗ f (x) + g(x)∗ f (x) dx. π]  f (0) = f (π) = 0}. f = 0.57) Note that the boundary conditions f (0) = f (π) = 0 are chosen such that the boundary terms occurring from integration by parts vanish. π) and consider the operator d2 f. it is straightforward to see ˜ that A is a nonnegative extension of A. the same calculation also shows that A is positive π 0 f (x)∗ (−f )(x)dx = 0 π f (x)2 dx > 0. In fact. Selfadjoint operators 59 (show this).
The converse is easy and hence d2 ˜ Af = − 2 f. Now deﬁne fn (x) = 0 gn (t)dt and note fn ∈ D(A) → f .60) Integration by parts on the right hand side shows π 0 g (x)∗ f (x)dx = − 0 π g (x)∗ 0 x ˜ (f (t) − f (t))dt dx (2. In other words. f A = g. we will consider 1 + A2 .63) Now let us apply this result to operators A corresponding to observables. π]  f (0) = f (π) = 0}.60 2. Moreover. not satisfy the assumptions of our lemma. So we see f ∈ H 2 (0. we infer Ran(A ± i) = H. f (0) = 0 is obvious π π and from 0 = fn (π) = 0 fn (t)dt we have f (π) = limn→∞ 0 fn (t)dt = 0. 1 + A2 is maximally deﬁned and hence is equal to this extension. Finally.9. So we have HA ⊆ {f ∈ H 1 (0. for any ϕ ∈ H there is a ψ ∈ D(A2 ) such that (A − i)(A + i)ψ = (A + i)(A − i)ψ = ϕ (2. Thus f ∈ AC[0. which has a symmetric extension whose range is H. let us compute the extension A. But there is another important consequence of the results which is worth while mentioning. dx ˜ D(A) = {f ∈ H 2 [0. f . Moreover. π)} and Af = −f . (2. By our requirement for observables.62) Now observe {g ∈ Hg ∈ HA } = {h ∈ H 0 h(t)dt = 0} = {1}⊥ and thus x ˜ f (x) + 0 (f (t) − f (t))dt ∈ {1}⊥⊥ = span{1}. Since A will. That is. ˜ all g ∈ HA there is an f π 0 x g (x)∗ f (x)dx = 0 π ˜ g(x)∗ (f (x) − f (x))dx. Selfadjointness and spectrum implies f (x) = 0 g(t)dt. We have f ∈ D(A) if for ˜ such that g. Observables correspond to selfadjoint operators. D(1 + A2 ) = D(A2 ).61) or equivalently π 0 g (x)∗ f (x) + 0 x ˜ (f (t) − f (t))dt dx = 0. As an immediate consequence we obtain Corollary 2. (2. in general. Ran(1 + A2 ) = H.64) and since (A ± i)ψ ∈ D(A). π (2. π)  f (0) = f (π) = 0}. . To see the converse π approximate f by smooth functions gn . π]. π]f ∈ H 1 (0. π) = ˜ {f ∈ AC[0. Using gn − 0 gn (t)dt instead of gn π x it is no restriction to assume 0 gn (t)dt = 0. instead. ˜ ˜ (iii).
it suﬃces to check that A − z is bijective. by γ and which satisﬁes D(A) Proof. Let A = − dx2 . By the closed graph theorem (Theorem 2. Problem 2.8. then BA is closed. so is A + z for any z ∈ C. (2. z ∈ ρ(A) if and only if (A − z) : D(A) → H is bijective and its inverse is bounded. Give an example where equality does not hold.) Problem 2. γ ∈ R. The rest is straightforward.7).4. that is.3. The resolvent set of A is deﬁned by ρ(A) = {z ∈ C(A − z)−1 ∈ L(H)}.8. 2 Problem 2. A2 ψ = 0. Show that A∗ A (with D(A∗ A) = {ψ ∈ D(A)Aψ ∈ D(A∗ )} is selfadjoint. Show that if A is bounded and B closed.65) ˜ Then there is a selfadjoint extension A which is also bounded from below ˜ ⊆ HA−γ . The complement of the resolvent set is called the spectrum σ(A) = C\ρ(A) (2. Show that if A is normal.3. Resolvents and spectra Let A be a (densely deﬁned) closed operator. z ∈ σ(A) if A − z has a nontrivial kernel. qA (ψ) = ψ. 2. A vector ψ ∈ Ker(A − z) is called an eigenvector and z is called eigenvalue in this case. Aψ = ψ. (Hint: A∗ A ≥ 0. Show that normal operators are closed.1. Problem 2.3. . Show (2. π)  f (0) = f (π) = 0} 1 and let ψ(x) = 2√π x(π −x).67) of A. Aψ ≥ γ ψ 2 . Problem 2. Find the error in the following argument: Since A is symmetric we have 1 = Aψ. Show that the kernel of a closed operator is closed.5.6.21). Show (αA)∗ = α∗ A∗ and (A+B)∗ ⊇ A∗ +B ∗ (where D(A∗ + B ∗ ) = D(A∗ ) ∩ D(B ∗ )) with equality if one operator is bounded.9. Suppose A is a closed operator. D(A) = {f ∈ H 2 (0.66) More precisely. In particular. Let A be a semibounded symmetric operator. Resolvents and spectra 61 Theorem 2.7. (2. d Problem 2. Problem 2. Problem 2.2.10 (Friedrichs extension). Problem 2. Show that A is normal if and only if AA∗ = A∗ A.2. Replacing A by A − γ we can reduce it to the case considered in Lemma 2.
Clearly (A − z)−1 is given by the multiplication operator (Af )(x) = A(x)f (x). Selfadjointness and spectrum The function RA : ρ(A) → L(H) z → (A − z)−1 is called resolvent of A. (2. 2π]  f ∈ L2 . But (A − z)−1 equivalent to µ({x A(x) − z ≥ ε}) = 0 and hence ≤ 1 ε is ρ(A) = {z ∈ C∃ε > 0 : µ({x A(x) − z ≤ ε}) = 0}. (Multiplication operator) Consider again the multiplication operator D(A) = {f ∈ L2 (Rn .73) Moreover. (2. Note the convenient formula RA (z)∗ = ((A − z)−1 )∗ = ((A − z)∗ )−1 = (A∗ − z ∗ )−1 = RA∗ (z ∗ ). (2. f (0) = f (2π)} (2.74) in L2 (0. By the variation of constants formula the solution is given by (this can also be easily veriﬁed directly) x (2. To compute the resolvent we must ﬁnd the solution of the corresponding inhomogeneous equation −if (x) − z f (x) = g(x).76) . Example.68) D((A − z)−1 ) = {f ∈ L2 (Rn . We already know that the eigenvalues of A are the integers and that the corresponding normalized eigenfunctions 1 un (x) = √ einx 2π form an orthonormal basis.62 2. (2. (Diﬀerential operator) Consider again the diﬀerential operator Af = −i d f. (A − z)−1 f (x) = 1 f (x). A(x) − z 1 f ∈ L2 (Rn . ρ(A∗ ) = ρ(A)∗ . dµ)  whenever this operator is bounded.70) Example. 2π). (2. z is an eigenvalue of A if µ(A−1 ({z})) > 0 and χA−1 ({z}) is a corresponding eigenfunction in this case. dµ)} (2.75) f (x) = f (0)eizx + i 0 eiz(x−t) g(t)dt. dx D(A) = {f ∈ AC[0. dµ)  Af ∈ L2 (Rn .71) given by multiplication with the measurable function A : Rn → C.69) In particular. dµ)}.72) A−z = 1 A−z ∞ (2.
ψ + zϕ) shows ϕ ∈ D(A) (since A is closed) and (A − z)R∞ ψ = ψ.2. Resolvents and spectra 63 Since f must lie in the domain of A. x. we have the ﬁrst resolvent formula RA (z) − RA (z ) = (z − z )RA (z)RA (z ) = (z − z )RA (z )RA (z).) Hence (A − z)−1 g(x) = 0 2π G(z. Then a quick calculation shows ARn ψ = ψ + (z − z0 )ϕn−1 + z0 ϕn . (2. Similarly. ϕ = R∞ ψ for some ψ ∈ H. Let R∞ = limn→∞ Rn and set ϕn = Rn ψ. (2.84) Hence (ϕn . (A − z)−1 − (z − z )(A − z)−1 (A − z )−1 = (A − z)−1 (1 − (z − A + A − z )(A − z )−1 ) = (A − z )−1 .83) converges to a bounded operator if z − z0  < RA (z0 ) −1 and clearly we expect z ∈ ρ(A) and Rn → RA (z) in this case. the inverse cannot exist in this case. x.3.79) (2. Thus R∞ = RA (z) as anticipated. The second follows after interchanging z and z . z ∈ ρ(A). we must have f (0) = f (2π) which gives f (0) = i e−2πiz − 1 0 2π e−izt g(t)dt.82) The sequence of bounded operators n Rn = j=0 (z − z0 )j RA (z0 )j+1 (2. In fact.80) RA (z) = j=0 (z − z0 )j RA (z0 )j+1 + (z − z0 )n+1 RA (z0 )n+1 RA (z). Rn Aψ = ψ + (z − z0 )ϕn−1 + z0 ϕn (2.78) where G(z. . If z. t) = eiz(x−t) In particular σ(A) = Z. 1−e2πiz t>x . 1−e−2πiz i . t<x z ∈ C\Z. (2. Aϕn ) → (ϕ. z ∈ C\Z. (2.77) (Since z ∈ Z are the eigenvalues. for ψ ∈ D(A). (2. Now ﬁx z = z0 and use (2.80) recursively to obtain n −i .85) and hence R∞ (A − z)ψ = ψ after taking the limit.81) which proves the ﬁrst equality. (2. t)g(t)dt.
z j+1 z > A . (2. In addition. If z is a boundary point of ρ(A).88) (2. RA (z) ≥ dist(z. Let ψn = RA (zn )ϕn and rescale ϕn such that ψn = 1. Let us also note the following spectral mapping result. Suppose z ∈ ρ(A)\{0}. then σ(A−1 )\{0} = (σ(A)\{0})−1 . by (2.11. that is. (2. Conversely. Let ψn be a Weyl sequence.91) . Selfadjointness and spectrum If A is bounded. then the converse is also true. Suppose A is injective.87) In fact. In addition. Then ϕn → 0 and hence (A − z)ψn = ϕn + (zn − z)ψn ≤ ϕn + z − zn  → 0 shows that ψn is a Weyl sequence. we have Aψ = zψ if and only if A−1 ψ = z −1 ψ. Proof. a similar argument veriﬁes the Neumann series for the resolvent n−1 RA (z) = − j=0 ∞ Aj 1 + n An RA (z) j+1 z z Aj .86) = − j=0 In summary we have proved the following Theorem 2. σ(A))−1 and if A is bounded we have {z ∈ C z > A } ⊆ ρ(A). We have z ∈ σ(A) if there is a sequence ψn ∈ D(A) such that ψn = 1 and (A − z)ψn → 0. it has an absolutely convergent power series expansion around every point z0 ∈ ρ(A). The resolvent set ρ(A) is open and RA : ρ(A) → L(H) is holomorphic.64 2. Lemma 2. Such a sequence is called Weyl sequence. Then we claim RA−1 (z −1 ) = −zARA (z) = −z − RA (z).87) there is a sequence zn → z and corresponding vectors ϕn ∈ H such that RA (z)ϕn ϕn −1 → ∞.89) (2. As a consequence we obtain the useful Lemma 2. ϕ ∈ H. (2.90) (2. Then z ∈ ρ(A) is impossible by 1 = ψn = RA (z)(A − z)ψn ≤ RA (z) (A − z)ψn → 0.13. Proof. the right hand side is a bounded operator from H → Ran(A) = D(A−1 ) and (A−1 − z −1 )(−zARA (z))ϕ = (−z + A)RA (z)ϕ = ϕ.12.
z ∈ C\(−∞. RA (z) ≤ Im(z)−1 and. Proof. ∞). 0]) and satisﬁes the given estimates as has been shown in the proof of Lemma 2. Then A is selfadjoint if and only if σ(A) ⊆ R and A ≥ 0 if and only if σ(A) ⊆ [0. if A ≥ 0. then Ran(A + z) = H. (2. then A is essentially selfadjoint. (2. z ∈ C\R.15. A ≥ 0). then A is essentially selfadjoint. Let A be selfadjoint.93) For the eigenvalues and corresponding eigenfunctions we have: Lemma 2. if A is selfadjoint (resp. Proof. Suppose A is a symmetric operator which has an orthonormal basis of eigenfunctions {ϕj }.6. λ < 0. In particular. Resolvents and spectra 65 Conversely. Theorem 2. let us characterize the spectra of selfadjoint operators. and hence A is selfadjoint by Lemma 2. .6. Aψ1 = ψ1 . Moreover. Then all eigenvalues are real and eigenvectors corresponding to diﬀerent eigenvalues are orthogonal. then inf σ(A) = inf ψ∈D(A). Next. if ψ ∈ D(A−1 ) = Ran(A) we have ψ = Aϕ and hence (−zARA (z))(A−1 − z −1 )ψ = ARA (z)((A − z)ϕ) = Aϕ = ψ. we obtain Theorem 2. However. However. If Aψj = λj ψj . 2. Aψ . Let A be symmetric. ψ2 − Aψ1 . Conversely. ψ2 = 0.92) Thus z −1 ∈ ρ(A−1 ). ψ =1 ψ. RA (λ) ≤ λ−1 . If H is ﬁnite dimensional. if there is an orthonormal basis of eigenvectors.3. then RA (z) exists for z ∈ C\R (resp.14.2. we can always ﬁnd an orthonormal basis of eigenvectors. j = 1. The rest follows after interchanging the roles of A and A−1 . Aψ1 = λ1 ψ1 .94) The result does not imply that two linearly independent eigenfunctions to the same eigenvalue are orthogonal. (2. Theorem 2. ψ1 = λ∗ ψ1 2 1 (2. Let A be symmetric. In the inﬁnite dimensional case this is no longer true in general. ψ2 = Aψ1 .95) 2 = ψ1 . we have λ1 ψ 1 and (λ1 − λ2 ) ψ1 . If σ(A) ⊆ R.17.16. λ1 ψ1 = ψ1 . it is no restriction to assume that they are since we can use GramSchmidt to ﬁnd an orthonormal basis for Ker(A − λ). it is essentially selfadjoint on span{ϕj }. ﬁnishing the proof. In particular.
U ∗ U ψ = ψ. What is the spectrum of an orthogonal projection? Problem 2. ˜ ˜ Write ψ = ψ + ϕ. U ψ2 = 0 since U ψ = zψ implies U ∗ψ = U −1 ψ = z −1 ψ = z ∗ ψ. (2. Consider the set of all ﬁnite linear combinations ψ = j=0 cj ϕj cj n which is dense in H. we can characterize the spectra of unitary operators. then σ(U ) ⊆ {z ∈ C z = 1}. If U ψj = zj ψj .96) In particular. ψ2 − ψ1 . Recall that a bijection U is called unitary if U ψ. Im(z) (2. D(A) = {f ∈ H 2 . U ψ = ψ.48). D(A) = {f ∈ H 1 [0. Selfadjointness and spectrum n Proof. Compute the resolvent of Af = f .19. where ψ ∈ D(A) and ϕ ≤ ε.11. U −1 is also unitary and hence σ(U ) ⊆ {z ∈ C z ≥ 1} by Lemma 2. ψ . In addition. Similarly. It suﬃces to prove the ﬁrst claim since the second then follows as in (2. (2. 2 (0. Compute the eigenvalues and eigenvectors of A = − dx2 . 2 we have (z1 − z2 ) ψ1 .98) Problem 2.99) Theorem 2. For every ψ ∈ H we have lim Im(z)→∞ ARA (z)ψ = 0.100) ˜ RA (z)Aψ + ARA (z)ϕ ˜ Aψ + ϕ . π)f (0) = f (π) = 0}. z j+1 z as Im(z) → ∞. if ψ ∈ D(An ). Then φ = j=0 λj ±i ϕj ∈ D(A) and (A ± i)φ = ψ shows that Ran(A ± i) is dense. Moreover. we note the following asymptotic expansion for the resolvent. ﬁnishing the proof. (2. then n RA (z)ψ = − j=0 1 Aj ψ + o( n+1 ). Let U be unitary.13. d Problem 2.18. Then ARA (z)ψ ≤ ≤ by (2.86). (2.12. Compute the resolvent of A. j = 1. Lemma 2. Thus U is unitary if and only if U ∗ = U −1 .10. ψ2 = U ∗ ψ1 .66 2. All eigenvalues have modulus one and eigenvectors corresponding to diﬀerent eigenvalues are orthogonal. 1]  f (0) = 0} and show that unbounded operators can have empty spectrum. Proof. Since U ≤ 1 we have σ(U ) ⊆ {z ∈ C z ≤ 1}.97) Proof. Suppose A is selfadjoint.
Find a Weyl sequence for the selfadjoint operator A = d2 − dx2 .. RA∗ A (z) = (A∗ RAA∗ (z)A − 1) . By Ran(A ± i) = (A ± i)D(A) = j (Aj ± i)D(Aj ) = j Hj = H we see that A is selfadjoint. if and only if both A1 and A2 are. The same considerations apply to countable orthogonal sums A= j Aj . 2. D(A) = H 2 (R) for z ∈ (0. (essentially) selfadjoint. Suppose Aj are selfadjoint operators on Hj . Let H1 ⊆ H be a closed subspace and let P1 be the corresponding projector. ∞). then RAj (z) ≤ ε−1 and hence −1 shows that z ∈ ρ(A). given an operator A it might be useful to write A as orthogonal sum and investigate each part separately. Orthogonal sums of operators 67 Problem 2. be two given Hilbert spaces and let Aj : D(Aj ) → Hj be two given operators.20. We say that H1 reduces the operator A if P1 A ⊆ AP1 . Orthogonal sums of operators Let Hj .104) where σ(A) = j σ(Aj ) (2.) Problem 2. then A = j Aj is selfadjoint and RA (z) = j RAj (z). Conversely.14. Suppose A is bounded. j RAj (z) ≤ ε Conversely. z z (2. Show that the spectrum of AA∗ and A∗ A coincide away from 0 by showing 1 1 RAA∗ (z) = (ARA∗ A (z)A∗ − 1) . D(A) = j D(Aj ) (2.105) (the closure can be omitted if there are only ﬁnitely many terms). z ∈ ρ(A) = C\σ(A) (2. Clearly A is closed. j σ(Aj )) > 0. j = 1.102) by setting A(ψ1 + ψ2 ) = A1 ψ1 + A2 ψ2 for ψj ∈ D(Aj ). What is σ(A)? (Hint: Cut oﬀ the solutions of −u (x) = z u(x) outside a ﬁnite ball.2. D(A) = D(A1 ) ⊕ D(A2 ) (2.101) 2. Note that this . Moreover. etc.4. Proof. Setting H = H1 ⊕ H2 we can deﬁne an operator A = A1 ⊕ A2 .103) and we have Theorem 2.13. if z ∈ σ(Aj ) there is a corresponding Weyl sequence ψn ∈ D(Aj ) ⊆ D(A) and hence z ∈ σ(A). if z ∈ j σ(Aj ) set ε = d(z.4.
However.106) If A is closable. In many physical applications a symmetric operator is given. P1 D(A) ⊆ D(A) and hence P2 D(A) = (I − P1 )D(A) ⊆ D(A). if ψ ∈ D(Aj ) we have Aψ = APj ψ = Pj Aψ ∈ Hj and thus Aj : D(Aj ) → Hj which proves the ﬁrst claim.108) 2. But this follows since ϕ. we have H = H1 ⊕H2 1 and P2 = I − P1 reduces A as well. suppose ψ ∈ D(A).5. Now let us turn to the second claim. Selfadjoint extensions It is safe to skip this entire section on ﬁrst reading. Hence selfadjointness is related to the dimension of these spaces and one calls the numbers d± (A) = dim K± . D(Aj ) = Pj D(A) ⊆ D(A).107) Proof. Selfadjointness and spectrum implies P1 D(A) ⊆ D(A). If A is selfadjoint. then A = A1 ⊕ A2 . As already noted. P1 Aψ2 = AP1 ϕ. Thus we see D(A) = D(A1 ) ⊕ D(A2 ). Since AP1 ψ = Aψ1 and P1 Aψ = P1 Aψ1 + P1 Aψ2 = Aψ1 + P1 Aψ2 we need to show P1 Aψ2 = 0.68 2. In particular. if it is not. Problem 2. Moreover. namely A. Moreover. then H1 also reduces A and A = A1 ⊕ A2 . In fact. if ψ ∈ D(A) we can write ψ = ψ1 ⊕ ψ2 .15. (2. Show (A1 ⊕ A2 )∗ = A∗ ⊕ A∗ . any other z ∈ C\R would be as good).21.2 we have seen that A is essentially selfadjoint if Ker(A∗ − z) = Ker(A∗ − z ∗ ) = {0} for one z ∈ C\R. If this operator turns out to be essentially selfadjoint. In Section 2. 1 2 (2. ψj = Pj ψ ∈ D(A). (2. But what happens in the general case? Is there more than . Then Pj ψn → Pj ψ and APj ψn = Pj Aψn → P Aψ. Pj ψ ∈ D(A) and APj ψ = P Aψ. Lemma 2. where Aj ψ = Aψ.109) defect indices of A (we have chosen z = i for simplicity. if we set H2 = H⊥ . Clearly A ⊆ A1 ⊕ A2 . (2. then there is a sequence ψn ∈ D(A) such that ψn → ψ and Aψn → Aψ. it is important to ﬁnd out if there are selfadjoint extensions at all (for physical problems there better are) and to classify them. If d− (A) = d+ (A) = 0 there is one selfadjoint extension of A. Conversely. K± = Ran(A ± i)⊥ = Ker(A∗ i). ψ2 = 0 for every ϕ ∈ D(A). Suppose H1 ⊆ H reduces A. then H1 reduces A if P1 D(A) ⊆ D(A) and AP1 ψ ∈ H1 for every ψ ∈ D(A). there is a unique selfadjoint extension and everything is ﬁne.
111) which shows that A is the Cayley transform of V and ﬁnishes the proof. if and only if d+ (A) = d− (A).110) Theorem 2.114) that is. Moreover. (2.116) .113) 2A(A + i)−1 . And thus for every ψ = (1 − V )ϕ ∈ D(A) = Ran(1 − V ) we have Aψ. let V be given and use the last equation to deﬁne A. A symmetric operator has selfadjoint extensions if and only if its defect indices are equal. that is. (1−V )ϕ = i (1−V )ϕ. Next observe 1 ± V = ((A − i) ± (A + i))(A + i)−1 = which shows D(A) = Ran(1 − V ) and A = i(1 + V )(1 − V )−1 . ﬁnding a selfadjoint extension of A is equivalent to ﬁnding a unitary extensions of V and this in turn is equivalent to (taking the closure and) ﬁnding a unitary operator from D(V )⊥ to Ran(V )⊥ . 2iV (1 − V )−1 (2. (2.115) (2. This is possible if and only if both spaces have the same dimension. ψ = −i (1+V )ϕ.23. In this case let A1 be a selfadjoint extension. Aψ . (1+V )ϕ = ψ. The Cayley transform is a bijection from the set of all symmetric operators A to the set of all isometric operators V (i.e. (1 V )ϕ = ±2iIm V ϕ. Finally observe A ± i = ((1 + V ) ± (1 − V ))(1 − V )−1 = 2i(1 − V )−1 . V ϕ = ϕ for all ϕ ∈ D(V )) for which Ran(1 + V ) is dense.112) (2. And thus for every ϕ = (A + i)ψ ∈ D(V ) = Ran(A + i) we have V ϕ = (A − i)ψ = (A + i)ψ = ϕ .2. Selfadjoint extensions 69 one extension. Theorem 2. Since A is symmetric we have (A ± i)ψ 2 = Aψ 2 + ψ 2 for all ψ ∈ D(A) by a straightforward computation. Thus A is selfadjoint if and only if its Cayley transform V is unitary. ϕ for all ϕ ∈ D(V ) by a straightforward computation. Then D(A1 ) = D(A) + (1 − V1 )K+ = {ψ + ϕ+ − V1 ϕ+ ψ ∈ D(A). 2i(A + i)−1 (2. Proof. ϕ+ ∈ K+ } (2.5. V1 its Cayley transform. Since V is isometric we have (1 ± V )ϕ.22.. A is symmetric. or maybe none at all? These questions can be answered by virtue of the Cayley transform V = (A − i)(A + i)−1 : Ran(A + i) → Ran(A − i). Conversely.
. (A1 ± i)−1 = (A ± i)−1 ⊕ where {ϕ+ } j i 2 j (2.121) e iθ = ˜ (2. fθ satisﬁes fθ (2π) = eiθ fθ (0). 1 − eiθ e2π e2π − eiθ The functions in the domain of the corresponding operator Aθ are given by (2. 2π)f (2π) = eiθ f (0)}. j j Corollary 2. Clearly K± = span{e x } is one dimensional and hence all unitary maps are of the form Vθ e2π−x = eiθ ex .122) (2.123) Concerning closures we note that a bounded operator is closed if and only if its domain is closed and any operator is closed if and only if its inverse is closed. D(A) = {f ∈ H 1 (0. α ∈ C. Hence d(z1 ) = d(z2 ) = d± (A).24. Proof. Moreover. K+ (z) = Ran(A + z)⊥ respectively K− (z) = Ran(A + z ∗ )⊥ . and thus we have D(Aθ ) = {f ∈ H 1 (0.117) ϕ± .70 2. j j (2. f ∈ D(A). 2π). ˜ ˜ (2. Hence we have Lemma 2.120) θ ∈ [0. 2π)f (0) = d f (2π) = 0} with adjoint A∗ = −i dx .25. First of all we note that instead of z = i we could use V (z) = (A + z ∗ )(A + z)−1 for any z ∈ C\R. Let d± (z) = dim K± (z). .118) is an orthonormal basis for K+ and ϕ− = V1 ϕ+ . Then dim Ker(A∗ − z ∗ ) = d for all z ∈ C\R. In particular. The same arguments as before show that there is a one to one correspondence between the selfadjoint extensions of A and the unitary operators on Cd(z) . • A is closed. D(A∗ ) = H 1 (0. The following items are equivalent.119) (2. d Example. • D(V ) = Ran(A + i) is closed. fθ (x) = f (x) + α(e2π−x − eiθ ex ). Recall the operator A = −i dx . (ϕ± − ϕj ). Selfadjointness and spectrum and A1 (ψ + ϕ+ − V1 ϕ+ ) = Aψ + iϕ+ + iV1 ϕ+ . 2π). Suppose A is a closed symmetric operator with equal defect indices d = d+ (A) = d− (A).
. Lemma 2.126) Proof. (2. Hence we have (A1 − z)−1 − (A2 − z)−1 = j ϕj (z). since D(A) = C 2 D(A) ⊆ CD(A). . (2.2. and ACψ = CAψ. ϕ ϕj (z) ∈ Ran(A − z)⊥ . . ϕj (z). An operator A is called Creal if CD(A) ⊆ D(A). If Aj . Finally. ψ ∈ D(A). (2. Cϕ = ψ. Let {ϕj } be an orthonormal set in Ran(A + i)⊥ . A skew linear map C : H → H is called a conjugation if it satisﬁes C 2 = I and Cψ.124) Note that in this case CD(A) = D(A).27. Then {Kϕj } is an orthonormal set in Ran(A − i)⊥ . ϕk (z ∗ ). (Al − z)−1 ϕk (z) .129) . ψj (z). Hence it suﬃces to consider it for vectors ϕ = j ϕj (z). (2.k 1 2 (αjk (z) − αjk (z)) ϕk (z). The prototypical example is of course complex conjugation Cψ = ψ ∗ . we give a useful criterion for the existence of selfadjoint extensions. Now computation the adjoint once using ((Aj − once using ( j ψj . ψj we obtain ϕj (z ∗ ).5. First observe that ((A1 − z)−1 − (A2 − z)−1 )ϕ is zero for every ϕ ∈ Ran(A−z). Hence the two spaces have the same dimension. then (A1 − z)−1 − (A2 − z)−1 = j.128) z ∗ )−1 and z)−1 )∗ = (Aj − ψj (z). Next. Suppose the symmetric operator A is Creal. . 2 are selfadjoint extensions and if {ϕj (z)} is an orthonormal basis for Ker(A∗ − z ∗ ). ϕj )∗ = j ϕj . (2.125) where l αjk (z) = ϕj (z ∗ ).26. Selfadjoint extensions 71 • Ran(V ) = Ran(A − i) is closed. then its defect indices are equal.127) where ψj (z) = ((A1 − z)−1 − (A2 − z)−1 )ϕj (z). • V is closed. we note the following useful formula for the diﬀerence of resolvents of selfadjoint extensions. ψj (z ∗ ) = j j (2. ϕ . . Theorem 2. Hence {ϕj } is an orthonormal basis for Ran(A + i)⊥ if and only if {Kϕj } is an orthonormal basis for Ran(A − i)⊥ . Proof. . j = 1.
b). In particular. b). Compute the defect indices of A0 = i dx . m ≥ 1. Appendix: Absolutely continuous functions Let (a.134) Then we have Lemma 2. b) = {f ∈ L2 (a. b)} ⊆ C[a. g ∈ L1 (a.130) and ﬁnishes the proof. Moreover. d ∞ Problem 2. ∞)).72 2. g is determined uniquely a. That is. 2. b)g ∈ L1 (a. b)} loc (2. By Corollary A. the limit is zero if the endpoint is inﬁnite. b] = {f ∈ AC(a. Proof..28. 0 ≤ j ≤ m − 1}.133) We set H (a. Note that AC(a. g ∈ AC[a. b). If the endpoint is inﬁnite.6.16.132) f (x)g (x) = f (b)g(b) − f (a)g(a) − a a f (x)g(x)dx. (2. note that x m f (j) (x)2 = f (j) (c)2 + 2 c Re(f (j) (t)∗ f (j+1) (t))dt (2. (with respect to Lebesgue measure) and f (x) = g(x). b) is a vector space. b) = {f ∈ C(a. We denote by x AC(a. then f (j+1) is integrable near this endpoint and hence the claim follows. b)f (j) ∈ AC(a. If f. If the endpoint is ﬁnite. Can you give a selfadjoint extension of A0 . b). D(A0 ) = Cc ((0. f is absolutely continuous if and only if it can be written as the integral of some locally integrable function. c ∈ (a.e. b) ⊆ R be some interval. f (j+1) ∈ L2 (a.135) shows that the limit exists (dominated convergence). ϕk (z) ϕj (z ∗ ) (2. Suppose f ∈ H m (a. b]. . b)f (x) = f (c) + c g(t)dt.e. Selfadjointness and spectrum Evaluating at ϕk (z) implies ψk (z) = j ψj (z ∗ ). (2.131) the set of all absolutely continuous functions. Then f is bounded and limx↓a f (j) (x) respectively limx↑b f (j) (x) exist for 0 ≤ j ≤ m − 1. b] we have the formula of partial integration b b x (2. If [a. b] is a compact interval we set AC[a. Since f (j) is square integrable the limit must be zero.33 f (x) = f (c) + c g(t)dt is diﬀerentiable a.
b)? (Hint: Start with the case where (a. What is the closure of C0 (a.1). Problem 2. b). b) = {f ∈ L2 (a.6.1 = a f (t)2 dt + a f (t)2 dt (2. (2.18. H m (a. f (r) ∈ L2 (a. the lower derivatives are then automatically in L2 .136) For a ﬁnite endpoint this is straightforward. b) in H 1 (a. b) is ﬁnite.2. Show that H 1 (a. For an inﬁnite endpoint this can also be shown directly. but it is much easier to use the Fourier transform (compare Section 7.19. that it suﬃces to check that the function plus the highest derivative is in L2 .17.137) ∞ Problem 2. (Hint: Fubini) Problem 2. b)f (j) ∈ AC(a. Appendix: Absolutely continuous functions 73 Let me remark. That is. 0 ≤ j ≤ m − 1. Show (2. 2 2. b) together with the norm b b f is a Hilbert space.) . b)}.133).
.
4) As long as f and g are polynomials. This matrix exponential can be deﬁned by a convergent power series ∞ (3. (f ∗ )(A) = f (A)∗ . (3.3) For this approach the boundedness of H is crucial. n! (3. that is. and H is hence a matrix.Chapter 3 The spectral theorem The time evolution of a quantum mechanical system is governed by the Schr¨dinger equation o d (3. no problems arise. we will need to perform some limiting procedure. is to diagonalize H. Hence we could consider convergent power series or equip the space of polynomials with the sup norm. which might not be the case for a a quantum system. dt If H = Cn .1. (f g)(A) = f (A)g(A). this system of ordinary diﬀerential equations is solved by the matrix exponential ψ(t) = exp(−itH)ψ(0).2) exp(−itH) = n=0 (−it)n n H . If we want to extend this deﬁnition to a larger class of functions.1) i ψ(t) = Hψ(t). the best way to compute the matrix exponential. But how do we diagonalize a selfadjoint operator? The answer is known as the spectral theorem. However. 3. such that (f +g)(A) = f (A)+g(A). and to understand the underlying dynamics. The spectral theorem In this section we want to address the problem of deﬁning functions of a selfadjoint operator A in a natural way. In both cases this only 75 .
. Pn ψ = Pn ψ 2 together with Lemma 1. Let H = Cn and A ∈ GL(n) be some symmetric matrix. Pn ψ = ψ. norm convergence does not even hold in the simplest case where H = L2 (I) and P (Ω) = χΩ (multiplication operator). the corresponding operators should be orthogonal projections. n P (Ωn )ψ = P (Ω)ψ.5) from the Borel sets to the set of orthogonal projections. then P (Ω)ψ for every ψ ∈ H (strong σadditivity).2). Then P (Ω) = χf −1 (Ω) (3.8) . λm be its (distinct) eigenvalues and let Pj be the projections onto the corresponding eigenspaces. To overcome this limitation. However. (ii) If Ω = n Ωn with Ωn ∩ Ωm = ∅ for n = m. Then PA (Ω) = {jλj ∈Ω} Pj (3. since wlim Pn = P for some orthogonal projections implies 2 slim Pn = P by ψ. . Since χΩ (λ)2 = χΩ (λ). rather than norm convergence. it even suﬃces to require weak convergence.11 (iv). (3. Moreover. n P (Ωn )ψ = Note that we require strong convergence. such that the following two conditions hold: (i) P (R) = I.76 3. In fact. . . A projectionvalued measure is a map P : B → L(H). that is. P (R\Ω) = I − P (Ω). Denote the Borel sigma algebra of R by B. we will defer this problem and begin by developing a functional calculus for a family of characteristic functions χΩ (A). Pn ψ = Pn ψ. Furthermore. (3.6) is a projection valued measure.7) is a projection valued measure (Problem 3. It is straightforward to verify that any projectionvalued measure satisﬁes P (∅) = 0. Let H = L2 (R) and let f be a realvalued measurable function. Ω → P (Ω). Let λ1 . P (Ω)∗ = P (Ω) and P (Ω)2 = P (Ω). Example. n P (Ωn ) = P (Ω). we will use characteristic functions χΩ (A) instead of powers Aj . The only remaining problem is of course j=1 the deﬁnition of χΩ (A). we should also have χR (A) = I and χΩ (A) = n χΩj (A) for any ﬁnite j=1 union Ω = n Ωj of disjoint sets. The spectral theorem works if the operator A is bounded. Example. since for a multiplication operator the norm is just the sup norm of the function.
Picking ψ ∈ H.16) we also have the following complex Borel measures 1 µϕ.ψ (Ω) = ϕ. Then. (ii) P (λ1 ) ≤ P (λ2 ) (that is ψ. The corresponding distribution function is given by µ(λ) = ψ. Moreover. P (λ1 )ψ ≤ ψ. strong right continuity is equivalent to weak right continuity.12) . Using the polarization identity (2.14) Note also that. suppose Ω1 ∩ Ω2 = ∅ ﬁrst. (3. (iv) slimλ→−∞ P (λ) = 0 and slimλ→+∞ P (λ) = I. P (λ2 )ψ ) or equivalently Ran(P (λ1 )) ⊆ Ran(P (λ2 )) for λ1 ≤ λ2 .9) Indeed. For the general case Ω1 ∩ Ω2 = ∅ we now have P (Ω1 )P (Ω2 ) = (P (Ω1 − Ω2 ) + P (Ω1 ∩ Ω2 ))(P (Ω2 − Ω1 ) + P (Ω1 ∩ Ω2 )) = P (Ω1 ∩ Ω2 ) as stated. (3.2). λ]) (3.ψ (Ω) ≤ ϕ ψ . we obtain a ﬁnite Borel measure µψ (Ω) = ψ. The spectral theorem 77 and P (Ω1 ∪ Ω2 ) + P (Ω1 ∩ Ω2 ) = P (Ω1 ) + P (Ω2 ). for every resolution of the identity there is a unique projection valued measure. we also have P (Ω1 )P (Ω2 ) = P (Ω1 ∩ Ω2 ).3): (i) P (λ) is an orthogonal projection.10) (3.11) Multiplying this equation from the right by P (Ω2 ) shows that P (Ω1 )P (Ω2 ) = −P (Ω2 )P (Ω1 )P (Ω2 ) is selfadjoint and thus P (Ω1 )P (Ω2 ) = P (Ω2 )P (Ω1 ) = 0. by CauchySchwarz.3. P (Ω)ψ = P (Ω)ψ 2 with µψ (R) = ψ 2 < ∞.1. To every projectionvalued measure there corresponds a resolution of the identity P (λ) = P ((−∞.13) which has the properties (Problem 3. P (λ)ψ and since for every distribution function there is a unique Borel measure (Theorem A. As before. (iii) slimλn ↓λ P (λn ) = P (λ) (strong right continuity). taking the square of (3. For any simple function f = n αj χΩj (where Ωj = f −1 (αj )) j=1 (3. Now let us turn to integration with respect to our projectionvalued measure.9) we infer P (Ω1 )P (Ω2 ) + P (Ω2 )P (Ω1 ) = 0. 4 (3. P (Ω)ψ = (µϕ+ψ (Ω) − µϕ−ψ (Ω) + iµϕ−iψ (Ω) − iµϕ+iψ (Ω)). µϕ.
Then ϕ. Recall that the set L(H) of all bounded linear mappings on H forms a C ∗ algebra.78 3. the operator P is a linear map from the set of simple functions into the set of bounded linear operators on H.17) Equipping the set of simple functions with the sup norm this implies P (f )ψ ≤ f ∞ ψ (3.19) . by linearity of the integral. (3.ψ (Ωj ) shows (3. P (f )ψ 2 = j αj 2 µψ (Ωj ) (the sets Ωj are disjoint) shows P (f )ψ 2 = R f (λ)2 dµψ (λ). if fn (x) → f (x) pointwise and if the sequence supλ∈R fn (λ) is bounded. For general f they follow from continuity. there is a unique extension of P to a bounded linear operator P : B(R) → L(H) (whose norm is one) from the bounded Borel functions on R (with sup norm) to the set of bounded linear operators on H. Theorem 3. Since the simple functions are dense in the Banach space of bounded Borel functions B(R).ψ (λ) and. that is. Moreover.18) that P has norm one. then P (fn ) → P (f ) strongly. P (f )ψ = ϕ.15) In particular. There is some additional structure behind this extension. The properties P (1) = I.16) and (3. and P (f g) = P (f )P (g) are straightforward for simple functions f . In particular. φ(ab) = φ(a)φ(b) and φ(a∗ ) = φ(a)∗ .16) f (λ)dµϕ.1. The spectral theorem we set n P (f ) ≡ R f (λ)dP (λ) = j=1 αj P (Ωj ). P (f ∗ ) = P (f )∗ . P (f )ψ = R αj µϕ. Proof. The last claim follows from the dominated convergence theorem and (3. j (3. A C ∗ algebra homomorphism φ is a linear map between two C ∗ algebras which respects both the multiplication and the adjoint. Then the operator P : B(R) → L(H) f → R f (λ)dP (λ) is a C ∗ algebra homomorphism with norm one. Hence P is a C ∗ algebra homomorphism.17) remain true. (3. Let P (Ω) be a projectionvalued measure on H. P (χΩ ) = P (Ω). (3.17). In addition.
26) holds for fn by our previous theorem. Since we expect the resulting operator to be unbounded. (3. For every ψ ∈ Df . ˜ ˜ P (fn )ψ converges to some vector ψ.16) and (3. Indeed.17). Next we want to deﬁne this operator for unbounded Borel functions. Theorem 3. dµψ ). we need a suitable domain ﬁrst. D(P (f )) = Df .ψ (λ) (3.24) converges to f in the sense of L2 (R.27) . One calls an unbounded operator A normal if D(A) = D(A∗ ) and Aψ = A∗ ψ for all ψ ∈ D(A).24) and abbreviate ψn = P (Ωn )ψ.23) This is clearly a linear subspace of H since µαψ (Ω) = α2 µψ (Ω) and since µϕ+ψ (Ω) ≤ 2(µϕ (Ω) + µψ (Ω)) (by the triangle inequality). dµψ ). P (f )ψ = P (f ∗ )ϕ. PA (f ) = A for f (λ) = λ. P (f )ψ = R g ∗ (λ)f (λ)dµϕ. the operator P (f ) = R f (λ)dP (λ). We deﬁne P (f )ψ = ψ. In addition.21) Example. Motivated by (3. Let f be given and deﬁne fn .25) is normal and satisﬁes P (f )∗ = P (f ∗ ).3. Df is dense. Ωn as above. By construction. Let H = Cn and A ∈ GL(n) respectively PA as in the previous example.1.2.17) hold. For any Borel function f .22) In particular.P (f )ψ = g ∗ f dµϕ. Moreover. (3. because of (3. ψn → ψ by (3.ψ . (3. P (f ) is a linear operator such that (3. (3. the bounded Borel function fn = χΩn f.17) since χΩn → 1 in L2 (R. Since (3.17) we set Df = {ψ ∈ H R f (λ)2 dµψ (λ) < ∞}. we get ϕ. (3. Ωn = {λ f (λ) ≤ n}. let Ωn be deﬁned as in (3. Now observe that dµψn = χΩn dµψ and hence ψn ∈ Df .26) Proof. observe P (g)ϕ. The spectral theorem 79 As a consequence. The operator P (f ) has some additional properties. (3. ψ (3.20) and thus dµP (g)ϕ. Moreover. Then m PA (f ) = j=1 f (λj )Pj .
P (f )ϕ = ψ.80 3. By (3. Moreover.31) Observe that this subspace is closed: If ψn = P (fn )ψ converges in H. ϕ for all ∗ )ψ = P (Ω )ψ since we have ˜ ϕ ∈ Df by deﬁnition. P (fn )ϕ = ψ. ψ . The vector ψ is called cyclic if Hψ = H. Now observe that P (fn n ∗ ˜ P (fn )ψ. The operators A in H and A in H are said to be unitarily equivalent if ˜ ˜ U A = AU. Thus it remains to show that ˜ D(P (f )∗ ) ⊆ Df . This proves existence of the limit n→∞ R lim ∗ fn 2 dµψ (λ) = lim P (fn )ψ n→∞ 2 ˜ = lim P (Ωn )ψ n→∞ 2 ˜ = ψ 2 . dµψ ) such that Uψ P (f ) = f Uψ . ψ ∈ Df = D(f ∗ ) by continuity. Luckily a spectral basis always exist: . U ψ = ˜ ˜ ϕ. U D(A) = D(A). dµψ )} ⊆ H. P (f )P (Ωn )ϕ = P (Ωn )ψ. dµψ ). Now let us return to our original problem and consider the subspace Hψ = {P (f )ψf ∈ L2 (R.28) for any ϕ ∈ H. (3. A is selfadjoint if and only if A is and σ(A) = σ(A). that is. If ψ is cyclic. (3. (3. ϕ = ψ. The spectral theorem for any ϕ. A set of spectral vectors is called a spectral basis if j Hψj = H. That P (f ) is normal follows from P (f )ψ 2 = P (f ∗ )ψ 2 = R f (λ)2 dµψ . Recall that U : H → H is called unitary if it is a bijection which preserves scalar products U ϕ. then fn converges to some f in L2 (since ψn − ψm 2 = fn − fm 2 dµψ ) and hence ψn → P (f )ψ. dµψ )f g ∈ L2 (R. ψ ∈ Df . if f is unbounded we have Uψ (Df ∩Hψ ) = D(f ) = {g ∈ L2 (R. dµψ )} (since ϕ = P (f )ψ implies dµϕ = f 2 dµψ ) and the above equation still holds.29) which implies f ∈ L2 (R. To see the second equality use P (fn )ϕ = P (fm χn )ϕ = P (fm )P (Ωn )ϕ for m ≥ n and let m → ∞.30) ˜ ˜ Clearly. our picture is complete.17). If ψ ∈ D(P (f )∗ ) we have ψ.33) (3. the relation Uψ (P (f )ψ) = f deﬁnes a unique unitary operator Uψ : Hψ → L2 (R. dµψ ). Otherwise we need to extend this approach.32) where f is identiﬁed with its corresponding multiplication operator. (3. ϕ (3. These considerations seem to indicate some kind of correspondence be˜ tween the operators P (f ) in H and f in L2 (R. A set {ψj }j∈J (J some index set) is called a set of spectral vectors if ψj = 1 and Hψi ⊥ Hψj for all i = j.
the spectrum is called simple. This can be easily done using a GramSchmidt type construction. since H is separable. it can be at most equal to the cardinality of an ordinary basis. Indeed. First of all observe that if {ψj }j∈J is a spectral set and ψ ⊥ Hψj for all j we have Hψ ⊥ Hψj for all j.4. g be Borel functions and α. take the orthogonal complement ψ1 . Normalize ψ1 and choose this to be ˜j which is not in Hψ . ψ = (1. However. Let H = C2 and A = . Choose the result to be ψ2 .36) Proof. Proceeding ˜ like this we get a set of spectral vectors {ψj } such that span{ψj } ⊆ j Hψj . The spectral theorem 81 Lemma 3. However.35) such that for any Borel function f . ˜ ˜ Now start with some total set {ψj }.3. P (g ∗ f )ψj = 0. ˜ Hence H = span{ψj } ⊆ Hψ . Using this canonical form of projection valued measures it is straightforward to prove Lemma 3. Then ψ1 = (1. there is an (at most countable) spectral basis {ψn } such that H= n Hψn (3. If A = there is no cyclic vector (why) 0 1 and hence the spectral multiplicity is two. ψ ⊥ Hψj implies P (g)ψ ⊥ Hψj for every bounded function g since P (g)ψ. it is at most countable. It suﬃces to show that a spectral basis exists. 0) and ψ2 = 0 1 (0. (3. U Df = D(f ). Move to the ﬁrst ψ 1 with respect to Hψ1 and normalize it. β ∈ C. D(αP (f ) + βP (g)) = Df +g (3. U P (f ) = f U. If the spectral multiplicity is one. The minimal cardinality of a spectral basis is called spectral multiplicity of P . Then we have αP (f ) + βP (g) ⊆ P (αf + βg). For every projection valued measure P . 1) are a spectral basis. j j It is important to observe that the cardinality of a spectral basis is not welldeﬁned (in contradistinction to the cardinality of an ordinary basis of the Hilbert space).34) and a corresponding unitary operator U= n Uψn : H → n L2 (R. dµψn ) (3. In particular. 0 0 Example.1. Let f.3.37) . But P (g)ψ with g bounded is dense in Hψ implying Hψ ⊥ Hψj . 1) is cyclic and hence the 1 0 spectrum of A is simple. P (f )ψj = ψ.
Such functions are called Herglotz functions. ψ ∈ H we can ﬁnd a corresponding complex measure µϕ. the measure µψ can be reconstructed from Fψ (z) by Stieltjes inversion formula 1 λ+δ Im(Fψ (t + iε))dt. λ−z (3.44) The measure µϕ. So by our above remarks. that is. if Fψ (z) is a Herglotz function satisfying F (z) ≤ Im(z) .ψ is conjugate linear in ϕ and linear in ψ. there is a corresponding measure µψ (λ) given by Stieltjes inversion formula. it is a Herglotz function.39) which is know as the Borel transform of the measure µψ .40) µψ (λ) = lim lim δ↓0 ε↓0 π −∞ M Conversely. By (3. for each ϕ. (3. Moreover.16) the corresponding quadratic form is given by Fψ (z) = ψ. So let A be a given selfadjoint operator and consider the expectation of the resolvent of A. Moreover. λ−z (3. Fψ (z) = ψ.43) that it maps the upper half plane to itself. D(P (f )P (g)) = Dg ∩ Df g . we consider the resolvent RA (z) = R (λ − z)−1 dP (λ). More generally.80) shows Im(Fψ (z)) = Im(z) RA (z)ψ 2 (3.42) (see Theorem 2. a comparison with our previous considerations begs us to deﬁne a family of . RA (z)ψ = R 1 dµϕ. (3.82 3. RA (z)ψ . To do this.14). The question is whether we can invert this map. The spectral theorem and P (f )P (g) ⊆ P (f g).38) Now observe. then it is the Borel transform of a unique measure µψ (given by Stieltjes inversion formula). the ﬁrst resolvent formula (2. (3. that to every projection valued measure P we can assign a selfadjoint operator A = R λdP (λ). by polarization. It is called spectral measure corresponding to ψ. RA (z)ψ = R 1 dµψ (λ). It can be shown (see Section 3. Moreover.ψ such that ϕ.41) This function is holomorphic for z ∈ ρ(A) and satisﬁes Fψ (z ∗ ) = Fψ (z)∗ and Fψ (z) ≤ ψ 2 Im(z) (3.ψ (λ).4) that Fψ (z) is a holomorphic map from the upper half plane to itself.
The relation PA (R) = I follows from (3.ψ (λ) = − dµϕ.48) by σadditivity of µψ .ψ ˜ 1 = ( ϕ.ψ (λ) = (3. Secondly we compute 1 dµ (λ) = ϕ. RA (z)PA (Ω)ψ = RA (z ∗ )ϕ. RA (˜)ψ ) z z−z ˜ 1 1 1 1 dµϕ. PA (Ω)ψ = R χΩ (λ)dµϕ. RA (z)RA (˜)ψ z z λ − z RA (z )ϕ. Now let Ω = n ∞ n=1 Ωn with Ωn ∩ Ωm = ∅ for n = m.93) below and Lemma 2.1. First observe (using the ﬁrst resolvent formula (2.47) R implying dµϕ. choosing Ω1 = Ω2 .5. Proof. (3. The operators PA (Ω) are non negative (0 ≤ ψ.PA (Ω)ψ 1 = χΩ (λ)dµRA (z ∗ )ϕ. PA (Ω)ψ λ − z ϕ.ψ (λ).ψ (Ω) ≤ ϕ ψ . We ﬁrst show PA (Ω1 )PA (Ω2 ) = PA (Ω1 ∩ Ω2 ) in two steps.46) z−z R λ−z λ−z ˜ ˜ ˜ R λ−z λ−z R implying dµRA (z ∗ )ϕ. PA (Ω1 ∩ Ω2 )ψ (3.8 since  ϕ. In particular. The family of operators PA (Ω) forms a projection valued measure. RA (z)ψ − ϕ. Lemma 3.ψ (λ) = χΩ (λ)dµϕ. The spectral theorem 83 operators PA (Ω) via ϕ. RA (˜)ψ = ϕ.ψ (λ). Then n ψ. Hence PA is weakly σadditive which implies strong σadditivity. PA (Ω)ψ  = µϕ. PA (Ω1 )PA (Ω2 )ψ = ϕ. PA (Ω)ψ = µψ (Ω) (3. .ψ (λ) R R λ−z ϕ.3. Equivalently we have since χΩ1 χΩ2 = χΩ1 ∩Ω2 .45) This is indeed possible by Corollary 1.18 which imply µψ (R) = ψ 2 .ψ (λ) = (λ − z)−1 dµϕ. PA (Ω)ψ ≤ 1) and hence selfadjoint. PA (Ωj )ψ = j=1 j=1 µψ (Ωj ) → ψ.ψ (λ) since a Herglotz function is uniquely determined by its measure.80)) 1 ∗ dµ (λ) = RA (z ∗ )ϕ. Now we can prove the spectral theorem for selfadjoint operators. we see that PA (Ω1 ) is a projector. as pointed out earlier.PA (Ω)ψ (λ) = χΩ (λ)dµϕ.
µ (3. then U RA (z) = RA (z)U and hence ˜ dµψ = d˜U ψ . set fε (λ) = χR\(λ0 −ε. ˜ ˜ Finally. U D(PA (f )) = D(PA (f )). that if A and A are unitarily equivalent as in (3.51) (which is larger than the domain D(A) = {ψ ∈ H R λ2 dµψ (λ) < ∞}). Conversely. Suppose PA (Ωn ) = 0. λ0 + ε)) = I.84 3. λ0 + n ). This extends our previous deﬁnition for nonnegative operators.30).54) we conclude λ0 ∈ σ(A) by Lemma 2. z ∈ C\R.7. Then (A − λ0 )PA (fε ) = PA ((λ − λ0 )fε (λ)) = PA (R\(λ0 − ε. we have U PA (f ) = PA (f )U .55) Similarly PA (fε )(A − λ0 ) = ID(A) and hence λ0 ∈ ρ(A). Let Ωn = (λ0 − n .50) and can be deﬁned for every ψ in the form domain selfadjoint operator Q(A) = {ψ ∈ H R λdµψ (λ) < ∞} (3.12. .ψ are uniquely determined by the resolvent and the projection valued measure is uniquely determined by the measures µϕ.ψ we are done. let us give a characterization of the spectrum of A in terms of the associated projectors. Then we can ﬁnd a ψn ∈ PA (Ωn )H with ψn = 1. (3. (3. Moreover. Existence has already been established. Since (A − λ0 )ψn 2 = = (A − λ0 )PA (Ωn )ψn R 2 (λ − λ0 )2 χΩn (λ)dµψn (λ) ≤ 1 n2 (3. The spectrum of A is given by σ(A) = {λ ∈ RPA ((λ − ε.4 shows that PA ((λ−z)−1 ) = RA (z).49) Proof.6 (Spectral theorem). if PA ((λ0 − ε.λ0 +ε) (λ)(λ − λ0 )−1 . Lemma 3.52) In particular. To every selfadjoint operator A there corresponds a unique projection valued measure PA such that A= R λdPA (λ). λ + ε)) = 0 for all ε > 0}. Theorem 3. λ0 + ε)) = 0. ˜ Note. The quadratic form of A is given by qA (ψ) = R λdµψ (λ) (3. (3. Since the measures µϕ. The spectral theorem Theorem 3.53) 1 1 Proof.
that is. dµ). In particular. √ Problem 3. 1}. (3. . Let A be a bounded operator and set √ A = A∗ A.4 (Polar decomposition).7) is a projection valued measure. Problem 3.1. Compute A = A∗ A of A = ϕ.2. This is the purpose of the present section. What is the corresponding operator? Problem 3. then this agrees with our present deﬁnition by Theorem 3. Show that Aψ = Aψ . Problem 3. where dµ is a ﬁnite Borel measure.56) (3. Show that P (λ) satisﬁes the properties (i)–(iv).57) PA ( j=0 αj λj ) = j=0 αj Aj . Show that (3.2. one needs to understand multiplication operators on L2 (R.3.3. PA (f ) is not aﬀected by the values of f on R\σ(A)! It is clearly more intuitive to write PA (f ) = f (A) and we will do so from now on.5.1. Problem 3.58) Moreover.1 showed that in order to understand selfadjoint operators. Conclude that Ker(A) = Ker(A) = Ran(A)⊥ and that U= ϕ = Aψ → Aψ ϕ→0 if ϕ ∈ Ran(A) if ϕ ∈ Ker(A) extends to a welldeﬁned partial isometry. we have the polar decomposition A = U A. In other words. . This notation is justiﬁed by the elementary observation n n and PA (R ∩ ρ(A)) = 0 (3.2. ψ. 3. U : Ker(U )⊥ → H is norm preserving. this also shows that if A is bounded and f (A) can be deﬁned via a convergent power series. More on Borel measures 85 Thus PA ((λ1 . Show that a selfadjoint operator P is a projection if and only if σ(P ) ⊆ {0. λ2 ) ⊆ ρ(A) and we have PA (σ(A)) = I and consequently PA (f ) = PA (σ(A))PA (f ) = PA (χσ(A) f ). λ2 )) = 0 if and only if (λ1 . More on Borel measures Section 3.
d(f µ)).61) By Theorem 3. it suﬃces to check this formula for simple functions g which follows since χΩ ◦ f = χf −1 (Ω) .63) (3.8.62) Now what can we say about the function f (A) (which is precisely the multiplication operator by f ) of A? We are only interested in the case where f is realvalued.66) It is tempting to conjecture that f (A) is unitarily equivalent to multiplication by λ in L2 (R. Lemma 3. λ + ε)) > 0 for all ε > 0}. (3. dµ) → L2 (R.67) However. (3. g → g ◦ f −1 (3. σ(µ) = {λ ∈ Rµ((λ − ε. g → g ◦ f. that is.65) In fact. dµ). then g(λ)d(f µ)(λ) = R R (3. that is. then U : L2 (R.f (λ) = g(λ)∗ f (λ)dµ(λ).59) is called the spectrum of µ. σ(A) = σ(µ). Introduce the measure (f µ)(Ω) = µ(f −1 (Ω)).60) is holomorphic for z ∈ C\σ(µ). (3.86 3. D(A) = {f ∈ L2 (R. dµ) is a cyclic vector for A and that dµg. Suppose f is injective. Invoking Morea’s together with Fubini’s theorem shows that the Borel transform F (z) = R 1 dµ(λ) λ−z (3. dµ). (3. The spectral theorem The set of all growth points. dµ)λf (λ) ∈ L2 (R.68) . In particular.64) g(f (λ))dµ(λ).7 the spectrum of A is precisely the spectrum of µ. this map is only unitary if its range is L2 (R. is a unitary map such that U f (λ) = λ. that is. Associated with this measure is the operator Af (λ) = λf (λ). The converse following from Stieltjes inversion formula. d(f µ)) → L2 (R. we have Pf (A) (Ω) = χf −1 (Ω) . (3. (3. d(f µ)) via the map L2 (R. dµ)}. Note that 1 ∈ L2 (R. if f is injective.
Moreover. σ(A) is compact. (f (A) − λ0 )gn < n−1 and hence λ0 ∈ σ(f (A)). where χ = χ(0. Note that we can still get a unitary map L2 (R. λ0 + n ) = ∅ without changing the corresponding operator.3. satisﬁes gn = 1. in addition. implying λ0 ∈ σ(f (A)). Ωn = 1 1 f −1 ((λ0 − n . If f is continuous. In particular.e.70) where equality holds if f is continuous and the closure can be dropped if. Then we also have U f (A1 ) = f (A2 )U for any bounded Borel Function and hence U f (λ) = U f (λ) · 1 = f (λ)U (1)(λ) (3. χd(f µ)) → L2 (R. σ(A) is bounded (i. dµ1 ) → L2 (R. More on Borel measures 87 Example. Let f be realvalued. Lemma 3. dµ2 ) be a unitary map such that U A1 = A2 U . Let A1 .∞) . then f (σ(A)) is compact and hence closed. If two operators with simple spectrum are unitarily equivalent can be read oﬀ from the corresponding measures: Lemma 3. then (g ◦ f )(λ) = g(λ2 ) and the range of the above map is given by the symmetric functions. Without restriction we can assume that Aj is multiplication by λ in L2 (R. since U is unitary we have µ1 (Ω) = R χΩ 2 dµ1 = R u χΩ 2 dµ2 = Ω u2 dµ2 . where v = U −1 1. f −1 (f (λ) − ε. The spectrum of f (A) is given by σ(f (A)) = σ(f µ). d(f µ)) ⊕ L2 (R. then the sequence gn = µ(Ωn )−1/2 χΩn . g2 ) → g1 (λ2 ) + g2 (λ2 )(χ(λ) − χ(−λ)). in addition. (3. dµ1 = u2 dµ2 . Proof. A2 be selfadjoint operators with simple spectrum and corresponding spectral measures µ1 and µ2 of cyclic vectors. Proof. Reversing the role of A1 and A2 we obtain dµ2 = v2 dµ1 .9. If λ0 ∈ σ(f µ). . (3. Let f (λ) = λ2 . dµj ). dµ). λ0 + n )). Thus (f (A) − λ0 )−1 = (f (λ) − λ0 )−1 exists and is bounded. Then A1 and A2 are unitarily equivalent if and only if µ1 and µ2 are mutually absolutely continuous. f (λ) + ε) contains an open interval around λ and hence f (λ) ∈ σ(f (A)) if λ ∈ σ(A). If. then µ(Ωn ) = 0 for some n and 1 1 hence we can change f on Ωn such that f (R) ∩ (λ0 − n . (g1 .10. Let U : L2 (R..72) that is.71) (3. compact). σ(f (A)) ⊆ f (σ(A)). if λ0 ∈ σ(f µ).69) and thus U is multiplication by u(λ) = U (1)(λ). Conversely.2.
Then σp (A) = { n n ∈ N} but σ(A) = σpp (A) = σp (A) ∪ {0}. where δn is the sequence which is 1 at the n’th place and zero else (that is. The spectral theorem The converse is left as an exercise (Problem 3.. σsc (A) = σ(µsc ). µs is supported. A is 1 1 a diagonal matrix with diagonal elements n ).) Next we recall the unique decomposition of µ with respect to Lebesgue measure. dµsc . singularly continuous. σac (A) = σ(µac ). (3. and Mpp .88 3. and pure point spectrum of A. Since the measures dµac . we have µac (B) = 0 for all B with Lebesgue measure zero) and µs is singular with respect to Lebesgue measure (i. dµsc ) ⊕ L2 (R.76) The corresponding spectra. and dµpp are mutually singular.. we have a corresponding direct sum decomposition of both our Hilbert space L2 (R.75) (3. We will choose them such that Mpp is the set of all jumps of µ(λ) and such that Msc has Lebesgue measure zero. To see this. (3.e. At z = 0 this formula still gives the inverse of A. respectively. (3. To the sets Mac . let H = 2 (N) and let A be given by Aδn = n δn . µs (R\B) = 0. (3. dµac ) ⊕ L2 (R. Since a continuous measure cannot live on a single point and hence also not on a countable set. Msc . just observe that δn is the 1 eigenvector corresponding to the eigenvalue n and for z ∈ σ(A) we have n RA (z)δn = 1−nz δn .77) . and Mpp correspond projectors P ac = χMac (A).e. but it is unbounded and hence 0 ∈ σ(A) but 0 ∈ σp (A). and P pp = χMpp (A) satisfying P ac + P sc + P pp = I. and σpp (A) = σ(µpp ) are called the absolutely continuous. Note that these sets are not unique. P sc = χMsc (A). The singular part µs can be further decomposed into a (singularly) continuous and a pure point part.74) where µsc is continuous on R and µpp is a step function. dµ) = L2 (R. they have mutually disjoint supports Mac . dµpp ) and our operator A A = (AP ac ) ⊕ (AP sc ) ⊕ (AP pp ). dµs = dµsc + dµpp . 1 Example. It is important to observe that σpp (A) is in general not equal to the set of eigenvalues σp (A) = {λ ∈ Rλ is an eigenvalue of A} since we only have σpp (A) = σp (A). on a set B with Lebesgue measure zero). we have σac (A) = σsc (A) = ∅. Msc . dµ = dµac + dµs .10.73) where µac is absolutely continuous with respect to Lebesgue measure (i. In other words.
8. Such a vector will be called a maximal spectral vector of A and µψ will be called a maximal spectral measure of A. t ∈ R. An example with purely singularly continuous spectrum is given by taking µ to be the Cantor measure. 3. Proof.9.10. Compute f µ. Spectral types Our next aim is to transfer the results of the previous section to arbitrary selfadjoint operators A using Lemma 3. But µψ (Ω) = j εj  µψj (Ω) = 0 implies µψj (Ω) = 0 for every j ∈ J and thus µϕ (Ω) = 0. Show the missing direction in the proof of Lemma 3.3. Determine the spectral types.1] (λ)dλ and f (λ) = χ(−∞. Problem 3. Let {ψj }j∈J be a spectral basis and choose nonzero numbers εj with 2 j∈J εj  = 1.10.78) d(f λ) = f (λ) Problem 3. then we can write it as ϕ = 2 2 j fj (A)ψj and hence dµϕ = j fj  dµψj . Problem 3. As in the unordered case one can show . then 1 dλ. Let dµ(λ) = χ[0. Construct a multiplication operator on L2 (R) which has dense point spectrum. Lemma 3. Show that if f ∈ AC(R) with f > 0. An example with purely absolutely continuous spectrum is given by taking µ to be Lebesgue measure.79) is a maximal spectral vector. (3.3.7. Then I claim that ψ= j∈J εj ψj (3.3. This will be the case if there is a vector ψ such that for every ϕ ∈ H its spectral measureµϕ is absolutely continuous with respect to µψ . Let A be the multiplication operator by the Cantor function in L2 (0. A set {ψj } of spectral vectors is called ordered if ψk is a maximal k−1 spectral vector for A restricted to ( j=1 Hψj )⊥ . Let λ be Lebesgue measure on R. 1).3. Let ϕ be given.6. Spectral types 89 Example. For every selfadjoint operator A there is a maximal spectral vector.t] . To do this we will need a spectral measure which contains the information from all measures in a spectral basis. Problem 3.11. Compute the spectrum of A. Problem 3.
in addition. ϕn = fn (A)ψn and hence dµϕn = fn 2 dµψn .75) for arbitrary selfadjoint operators A. In particular.85) .xx . pp}. we want to introduce the splitting (3. We will use the unitary operator U of Lemma 3. It is tempting to pick a spectral basis and treat each summand in the direct sum separately. For the sets Mxx one can choose the corresponding supports of some maximal spectral measure µ.84) and hence dµϕ. Hsc = {ψ ∈ Hµψ is singularly continuous}. σ(A) is bounded. (3. If µ is a maximal spectral measure we have σ(A) = σ(µ) and the following generalization of Lemma 3.80) (3. We have H = Hac ⊕ Hsc ⊕ Hpp . Theorem 3. the subspaces Hxx reduce A. For every selfadjoint operator there is an ordered spectral basis. sc. Then the spectrum of f (A) is given by σ(f (A)) = {λ ∈ Rµ(f −1 (λ − ε. Observe that if {ψj } is an ordered spectral basis.82) fn 2 dµψn (3. Let µ be a maximal spectral measure and let f be realvalued. xx ∈ {ac. Lemma 3. λ + ε)) > 0 for all ε > 0}. then. then µψj+1 is absolutely continuous with respect to µψj . Hpp = {ψ ∈ Hµψ is pure point}. The spectral theorem Theorem 3. since it is not clear that this approach is independent of the spectral basis chosen.14.83) There are Borel sets Mxx such that the projector onto Hxx is given by P xx = χMxx (A).81) where equality holds if f is continuous and the closure can be dropped if. Pick ϕ ∈ H and write ϕ = n ϕn with ϕn ∈ Hψn . Next.xx = n fn 2 dµψn . Let fn = U ϕn .13 (Spectral mapping). σ(f (A)) ⊆ f (σ(A)).12. we use the more sophisticated deﬁnition Hac = {ψ ∈ Hµψ is absolutely continuous}.9 holds. pp}. (3. However.90 3.3. Proof. by construction of the unitary operator U . since the subspaces Hψn are orthogonal. xx ∈ {ac. In particular. sc. we have dµϕ = n (3. Moreover. (3.
singularly continuous. and pure point spectrum of A are deﬁned as and σpp (A) = σ(AHpp ). Compute σ(A).52). λ2 ])) = lim ε↓0 π 2 λ2 Im(F (λ + iε))dλ. Problem 3. In particular. .89) F (z) ≤ Im(z) Moreover. σxx (A) = σxx (A).4. dµψn . We can deﬁne F on C− using F (z ∗ ) = F (z)∗ .86) and reduces our problem to the considerations of the previous section. z ∈ C+ .xx ). and σpp (A) for the multi1 plication operator A = 1+x2 in L2 (R).87) respectively.3. The absolutely continuous. pp} (3. Stieltjes inversion formula follows from Fubini’s theorem and the dominated convergence theorem since 1 2πi λ2 ( λ1 1 1 1 − )dλ → χ (x) + χ(λ1 . What is its spectral multiplicity? σac (A) = σ(AHac ). and σpp (A) = σ(µpp ). The Borel transform of a ﬁnite Borel measure is a Herglotz function satisfying µ(R) . Then its Borel transform is deﬁned via F (z) = R dµ(λ) . λ−z (3. If µ is a maximal spectral measure we have σac (A) = σ(µac ). xx ∈ {ac. (3. Appendix: The Herglotz theorem A holomorphic function F : C+ → C+ . σsc (A) = σ(µsc ). σsc (A). 3.11. Suppose µ is a ﬁnite Borel measure.90) Proof.λ2 ] pointwise.4.λ2 ) (x) (3.88) Theorem 3. By Morea’s and Fubini’s theorem. is called a Herglotz function. the measure µ can be reconstructed via Stieltjes inversion formula 1 1 (µ((λ1 . (3. ˜ ˜˜ If A and A are unitarily equivalent via U . λ2 )) + µ([λ1 . sc. F is holomorphic on C+ and the remaining properties follow from 0 < Im((λ−z)−1 ) and λ−z−1 ≤ Im(z)−1 .15. Appendix: The Herglotz theorem 91 This shows U Hxx = n L2 (R. λ1 (3. σsc (A) = σ(AHsc ). σac (A).91) x − λ − iε x − λ + iε 2 [λ1 . C± = {z ∈ C ± Im(z) > 0}. then so are AHxx and AHxx ˜ by (3.
100) In particular.92) (3. (3. such that F is the Borel transform of µ. Suppose F is a Herglotz function satisfying M F (z) ≤ . satisfying µ(R) ≤ M . Hence we have by Cauchy’s formula F (z) = 1 2πi Γ 1 1 − ∗ − 2iε ζ −z ζ −z F (ζ)dζ. The spectral theorem Observe Im(F (z)) = Im(z) R dµ(λ) λ − z2 (3. The converse of the previous theorem is also true Theorem 3. Proof.93) and λ→∞ lim λ Im(F (iλ)) = µ(R). We abbreviate F (z) = v(z) + i w(z) and z = x + i y.92 3. z ∈ C+ .95) and note that z lies inside Γ and z ∗ + 2iε lies outside Γ if 0 < ε < y < R.96) Inserting the explicit form of Γ we see F (z) = 1 π y−ε F (x + iε + λ)dλ λ2 + (y − ε)2 −R i π y−ε + F (x + iε + Reiϕ )Reiϕ dϕ. Im(z) (3. R (3.99) where φε (λ) = (y − ε)/((λ − x)2 + (y − ε)2 ) and wε (λ) = w(λ + iε)/π. Next we choose a contour Γ = {x + iε + λλ ∈ [−R. 2 e2iϕ + (y − ε)2 π 0 R R (3.16. since φε (λ) − φ0 (λ) ≤ const ε we have w(z) = lim ε↓0 R φ0 (λ)dµε (λ). (3.94) Then there is a unique Borel measure µ. π]}.98) π R (λ − x)2 + (y − ε)2 and taking imaginary parts w(z) = R φε (λ)wε (λ)dλ.97) The integral over the semi circle vanishes as R → ∞ and hence we obtain 1 y−ε F (z) = F (λ + iε)dλ (3.101) . (3. Letting y → ∞ we infer from our bound wε (λ)dλ ≤ M. R]} ∪ {x + iε + Reiϕ ϕ ∈ [0. (3.
λm+1 ≥ x + . (3.102) Fix δ > 0 and let λ1 < λ2 < · · · < λm+1 be rational numbers such that λj+1 − λj  ≤ δ and λ1 ≤ x − y3 y3 . and we claim w(z) = R λ φ0 (λ)dµ(λ). The RadonNikodym derivative of µ can be obtained from the boundary values of F . Now F (z) and R (λ − z)−1 dµ(λ) have the same imaginary part and thus they only diﬀer by a real constant. n ≥ N.106) can be made arbitrarily small. there is even a subsequence εn such that µεn (λ) converges for each rational λ. (3. 0 ≤ µ(λ1 ) ≤ µ(λ2 ) ≤ M .104) δ . In summary.106) The ﬁrst and third term can be bounded by 2M δ/y 2 . Moreover. Moreover.105) φ0 (λ)dµ(λ) − R φ0 (λ)dµεn (λ) ≤ m  R φ0 (λ)dµ(λ) − j=1 m φ0 (λj )(µ(λj+1 ) − µ(λj )) + j=1 φ0 (λj )(µ(λj+1 ) − µ(λj ) − µεn (λj+1 ) + µεn (λj )) m + R φ0 (λ)dµεn (λ) − j=1 φ0 (λj )(µεn (λj+1 ) − µεn (λj )) (3. Since 0 ≤ µε (λ) ≤ M . δ δ (3. For irrational λ we set µ(λ0 ) = inf λ≥λ0 {µ(λ)λ rational}. By our bound this constant must be zero. . Appendix: The Herglotz theorem 93 where µε (λ) = −∞ wε (x)dx. It remains to establish that the monotone functions µε converge properly. (3. by the standard diagonal trick.4. Then µ(λ) is monotone.3.103) Then (since φ0 (λ) ≤ y −2 ) φ0 (λ) − φ0 (λj ) ≤ and φ0 (λ) ≤ Now observe  R δ . λ1 ≤ λ2 . (3.107) 2m and hence the second term is bounded by δ. there is a convergent subsequence for ﬁxed λ. y2 λj ≤ λ ≤ λj+1 . the diﬀerence in (3. since φ0 (y) ≤ 1/y we can ﬁnd an N ∈ N such that y µ(λj ) − µεn (λj ) ≤ δ. y2 λ ≤ λ1 or λm+1 ≤ λ.
35 we obtain Theorem 3. (3. t − λ < s < δ} and obtain δ µ(Is )Kε (s)ds = 0 Iδ (K(δ) − Kε (t − λ))dµ(t).111) To estimate the ﬁrst part we integrate Kε (s) ds dµ(t) (3. (3. Let µ be a ﬁnite Borel measure and F its Borel transform. (3.112) over the triangle {(s. . t)λ − δ < t < λ + δ.110) Clearly the second part can be estimated by Kε (t − λ)µ(t) ≤ Kε (δ)µ(R). Kε (t) = t2 ε . Iδ = (λ−δ.115) As a consequence of Theorem A.18. (3. then δ δ 2c arctan( ) ≤ Kε (t − λ)dµ(t) ≤ 2C arctan( ) ε ε Iδ since δKε (δ) + δ −sKε (s)ds = arctan( ). then 1 1 (Dµ)(λ) ≤ lim inf F (λ + iε) ≤ lim sup F (λ + iε) ≤ (Dµ)(λ). λ+δ). The spectral theorem Theorem 3. Let µ be a ﬁnite Borel measure and F its Borel transform. 0 < s < δ} = {(s. t)λ − s < t < λ + s.17. R\Iδ (3.94 3. δ ≤ C. then the limit 1 (3.109) We ﬁrst split the integral into two parts Im(F (λ+iε)) = Iδ Kε (t−λ)dµ(t)+ R\Iδ Kε (t−λ)µ(t).113) µ(Is ) 2s Now suppose there is are constants c and C such that c ≤ 0 ≤ s ≤ δ.116) Im(F (λ)) = lim Im(F (λ + iε)) ε↓0 π exists a.e.114) (3. ε 0 Thus the claim follows combining both estimates. with respect to both µ and Lebesgue measure (ﬁnite or inﬁnite) and 1 (Dµ)(λ) = Im(F (λ)) (3.117) π whenever (Dµ)(λ) exists. We need to estimate Im(F (λ + iε)) = R Kε (t)dµ(t).108) ε↓0 π π ε↓0 Proof.34 and Theorem A. + ε2 (3.
Appendix: The Herglotz theorem 95 Moreover. .19. The measure µ is purely absolutely continuous on I if lim supε↓0 Im(F (λ + iε)) < ∞ for all λ ∈ I. the set {λF (λ) = ∞} is a support for the singularly and {λF (λ) < ∞} is a support for the absolutely continuous part. Corollary 3.4.3. In particular.
.
Integral formulas We begin with the ﬁrst task by having a closer look at the projector PA (Ω). In particular. which are known to hold for the resolvent of an operator A. the number ψ. χΩ (A)ψ (4. Moreover. 4. We will give a few typical applications: Firstly we will derive an operator valued version of of Stieltjes’ inversion formula. Secondly we will consider commuting operators and show how certain facts. we need to show how to integrate a family of functions of A with respect to a parameter. Now let us show how the spectral theorem can be used.1. To do this. we will investigate tensor products of operators. can be established for a larger class of functions.Chapter 4 Applications of the spectral theorem This chapter can be mostly skipped on ﬁrst reading. Finally. we will show that these integrals can be evaluated by computing the corresponding integrals of the complex valued functions. You might want to have a look at the ﬁrst section and the come back to the remaining ones later. They project onto subspaces corresponding to expectation values in the set Ω. Then we will show how the eigenvalues below the essential spectrum and dimension of RanPA (Ω) can be estimated using the quadratic form.1) 97 .
5) Similarly.8) . where Ω denotes the Lebesgue measure of Ω. (4.ψ (λ) = λ0 ϕ. consider the simple function n−1 t1 −t0 fn = i=0 f (si )χ[si . Observe that C(I. ψ (4. ε↓0 (4. X) is uniformly continuous.ψ (λ) = λ0 R dµϕ. Moreover. ψ = 1. 1 ≤ i ≤ n. In fact. Aψ = R λ χ{λ0 } (λ)dµϕ. The space Ranχ{λ0 } (A) is called the eigenspace corresponding to λ0 since we have ϕ. we can obtain an operator valued version of Stieltjes’ inversion formula. X) → X by (4. X) we can deﬁne a linear map n : S(I. For f ∈ S(I. But ﬁrst we need to recall a few facts from integration in Banach spaces. Applications of the spectral theorem is the probability for a measurement of a to lie in Ω.6) The corresponding Banach space obtained after completion is called the set of regulated functions R(I. since −iε = χ{λ0 } (λ) ε↓0 λ − λ0 − iε we infer from Theorem 3.7) f (t)dt = I i=1 xi f −1 (xi ). In addition. X). The dimension of the eigenspace is called the multiplicity of the eigenvalue.2) where hull(Ω) is the convex hull of Ω. a function f : I → X is called simple if the image of f is ﬁnite.98 4. is a Borel set.1 that lim lim −iεRA (λ0 + iε)ψ = χ{λ0 } (A)ψ. t1 ] ⊂ R is a compact interval and X is a Banach space. X) forms a linear space and can be equipped with the sup norm f ∞ = sup f (t) . The set of simple functions S(I. Aψ = Ω λ dµψ (λ) ∈ hull(Ω). and if each inverse i=1 image f −1 (xi ).si+1 ) . we have ψ. Since f ∈ C(I. where si = t0 + i n . ψ ∈ PA (Ω)H. We will consider the case of mappings f : I → X where I = [t0 . X). we infer that fn converges uniformly to f .3) and hence Aψ = λ0 ψ for all ψ ∈ Ranχ{λ0 } (A).4) (4. X) ⊂ R(I. As before. f (I) = {xi }n . This map satisﬁes f (t)dt ≤ f I ∞ (t1 − t0 ) (4. t∈I (4.
Integral formulas 99 and hence it can be extended uniquely to a linear map : R(I. Lemma 4. λ)dt.t3 ) (s)f (s)ds and f (s)ds.. f ∈ R(I. if f ∈ C(I. L(H)). (4.11) If I = R. In this case we can set f (t)dt = lim R r→∞ [−r. then A(t)dt ψ = I I (A(t)ψ)dt. . X) und thus for all f ∈ R(I. λ) and set F (λ) = I f (t.24. A)dt respectively F (A)ψ = I f (t.r] f (t)dt (4.12) and (4. (4. A) ∈ R(I. then F (t) = C 1 (I. A)ψ dt.15) Proof. Suppose f : I × R → C is a bounded Borel function such that f (.9) and (4. if ∈ X ∗ is a continuous linear functional. We even have f (t)dt ≤ I I f (t) dt.10) In particular. if A(t) ∈ R(I. A) ∈ R(I. L(H)) follows from the spectral theorem. Then f (t. X) for all r > 0 and if f (t) is integrable.t+ε] (4. We will use the standard notation t2 t3 f (s)ds = − t3 t2 t3 t2 f (s)ds = I χ(t2 . X). then ( I f (t)dt) = I (f (t))dt. f (t + ε) − f (t) d f (t) = lim ε→0 dt ε We write f ∈ C 1 (I. Let A be selfadjoint. since it is no restriction to assume that A is multiplication by λ in some L2 space. X). s∈[t. we say that f : I → X is integrable if f ∈ R([−r. X) by continuity. In particular. X) and dF/dt = f as can be seen from t+ε f (s)ds ∈ F (t + ε) − F (t) − f (t)ε = t (f (s) − f (t))ds ≤ ε sup f (s) − f (t) . In addition.4.9) which clearly holds for f ∈ S(I.1. (4. X) → X with the same norm (t1 − t0 ) by Theorem 0.10) still hold. r]. (4.1. X) if (4. That f (t.14) The important facts for us are the following two results.13) t t0 exists for all t ∈ I. L(H)) and F (A) = I f (t.
Show χΩ (A) = Γ RA (z)dz.17) Proof. Problem 4.20) λ−z for one z ∈ C\K is dense in C∞ (K). λ)dt dµϕ. It suﬃces to prove the case where f is simple and of compact support. Then A R f (t)dt = R Af (t)dt respectively R f (t)dtA = R f (t)Adt.3 (Stone’s formula).1 with Theorem 3.ψ (λ) F (λ)dµϕ. F (A)ψ R = (4. A)ψ dt f (t. A)dt)ψ I = I ϕ. But for such functions the claim is straightforward. Proof. ( f (t.2. (4. λ)dµϕ. λ2 ]) + PA ((λ1 . (4. λ2 ))) 2 (4. Lemma 4. . Let K ⊆ R be closed.1. The ∗algebra generated by the function 1 λ→ (4. Let Γ be a diﬀerentiable Jordan curve in ρ(A). The result follows combining Lemma 4.16) by Fubini’s theorem and hence the ﬁrst claim follows.ψ (λ) = ϕ.91). Suppose f : R → L(H) is integrable and A ∈ L(H).18) strongly.19) where Ω is the intersection of the interior of Γ with R. And let C∞ (K) be the set of all continuous functions on K which vanish at ∞ (if K is unbounded) with the sup norm.1 and (3. then 1 2πi λ2 (RA (λ + iε) − RA (λ − iε))dλ → λ1 1 (PA ([λ1 .2. Applications of the spectral theorem We compute ϕ. Theorem 4.100 4. Now we can prove Stone’s formula. Commuting operators Now we come to commuting operators. 4. f (t. As a preparation we can now prove Lemma 4.ψ (λ)dt I R = = R I f (t.4. Let A be selfadjoint.
Hence if A is selfadjoint and B is bounded we will say that A and B commute if [RA (z). n→∞ n→∞ (4. Otherwise. dµψ ). Choose a sequence fn ∈ C∞ (σ(A)) converging to f in L2 (R. Equation (4.25) shows f ∈ L2 (R. Since σ(A) is compact. We say that two bounded operators A. B] = AB − BA = 0.23) by our lemma. . Next ﬁx ψ ∈ H and let f be bounded. Conversely. the claim follows directly from the complex Stone– Weierstraß theorem since (λ1 −z)−1 = (λ2 −z)−1 implies λ1 = λ2 . If f is unbounded. which is compact. B commute if [A. A good candidate is the resolvent.e. Proof. B] = [RA (z ∗ ). Commuting operators 101 Proof.2. the claim holds for any ψ ∈ D(f (A)).22) (4. (4. Suppose A is selfadjoint and commutes with the bounded operator B. If A is selfadjoint and bounded.4.g. Then Bf (A)ψ = lim Bfn (A)ψ = lim fn (A)Bψ = f (A)Bψ. Then f (A)Bψ = lim fn (A)Bψ = lim Bfn (A)ψ n→∞ n→∞ (4. the claim follows for all such f ∈ C∞ (σ(A)). If K is compact. ˜ replace K by K = K ∪{∞}. To avoid this nuisance we will replace A by a bounded function of A. dµBψ ) (i.23) for any bounded Borel function f . Then [f (A).6. B] = 0 (4. and set (∞−z)−1 = 0.21) follows from (4. Lemma 4. Since this subalgebra is dense in C∞ (σ(A)). B] = 0 for one z ∈ ρ(A). since B commutes with any polynomial of A.24). we soon run into trouble with this deﬁnition since the above expression might not even make sense for any nonzero vector (e.24) If f is unbounded. take B = ϕ.22) tell us that (4.21) If A or B is unbounded.. ψ with ψ ∈ D(A)).5.23) holds for any f in the ∗subalgebra generated by RA (z). Corollary 4. let ψ ∈ D(f (A)) and choose fn as in (3. we have λ ∈ C∞ (σ(A)) and hence (4. the claim follows from the Neumann series. Proof.. As another consequence we obtain . then (4.21) holds.22) holds if and only if (4. Then we can again apply the complex Stone–Weierstraß theorem to conclude that ˜ our ∗subalgebra is equal to {f ∈ C(K)f (∞) = 0} which is equivalent to C∞ (K). Bψ ∈ D(f (A))) and f (A)Bψ = BF (A)ψ.
The assumption that the spectrum of A is simple is crucial as the example A = I shows.1.7.8. Then Bg(λ) = Bg(λ) · 1 = g(λ)(B1)(λ) (4. Let ψ be a cyclic vector for A. Hence B is multiplication by f (λ) = (B1)(λ). e−iBs ] = 0. A bounded operator B commutes with A if and only if B = f (A) for some bounded Borel function. since these functions ˆ are dense in C∞ (R) by the complex Stone–Weierstraß theorem.31) . From our above analysis it follows that this is equivalent to [e−iAt . see Section 7. B]dt = 0 (4. B] = √ 2π R f (t)[e−iAt . Proof. B] = √ [ 2π by Lemma 4. respectively t. Suppose A is selfadjoint and has simple spectrum. Lemma 4. Note also that the functions exp(−itA) can also be used instead of resolvents. Here f denotes the Fourier transform of f .29) ∗ for one z1 ∈ ρ(A) and one z2 ∈ ρ(B) (the claim for z2 follows by taking adjoints). It suﬃces to show [f (A). (4. RB (z2 )] = 0 (4. s ∈ R. By our unitary equivalence it is no restriction to assume H = L2 (R. (4. But for such f we have 1 ˆ [f (A). B] = 0 for f ∈ S(R). g(B)] = 0 for arbitrary bounded Borel functions f and g. Then B commutes with A if and only if [e−iAt . dµψ ). The extension to the case where B is selfadjoint and unbounded is straightforward. Suppose A is selfadjoint and B is bounded.2.26) since B commutes with the multiplication operator g(λ). B] = 0 for all t ∈ R. Applications of the spectral theorem Theorem 4.30) [f (A). We say that A and B commute in this case if ∗ [RA (z1 ).28) R (4.27) 1 f (t)e−iAt dt. RB (z2 )] = [RA (z1 ). ˆ Proof.102 4.
. The minmax theorem 103 4. Aψ ≤ En + O(ε).4. Deﬁne U (ψ1 . If we restrict to the orthogonal complement of an approximating eigenfunction ψ1 . In general there is no way of computing the lowest eigenvalues and their corresponding eigenfunctions explicitly. ψn−1 ). . ψn ) = {ψ ∈ D(A) ψ = 1. .35) and the claim follows. So suppose we have a normalized function ψ1 which is an approximation for the eigenfunction ϕ1 of the lowest eigenvalue E1 ..3. The minmax theorem In many applications a selfadjoint operator has a number of eigenvalues below the bottom of the essential spectrum. The essential spectrum is obtained from the spectrum by removing all discrete eigenvalues with ﬁnite multiplicity (we will have a closer look at it in Section 6. .ψn−1 ) (4. Assume they satisfy (A − Ej )ϕj = 0. ψn }⊥ }. .15 we know that ψ1 . . (i) We have inf ψ∈U (ψ1 .33) ψ. ..2).3. then we can restrict A to the orthogonal complement of ϕ1 and proceed as before: E2 will be the inﬁmum over all expectations restricted to this subspace. Aψ1 ≥ ϕ1 . . (4. let {ϕj }N be an orthonormal basis for the space spanned j=1 by the eigenfunctions corresponding to eigenvalues below the essential spectrum. set ψ = then n j=1 αj ϕj and choose αj such that ψ ∈ U (ψ1 . Aψ = j=1 αj 2 Ej + O(ε) ≤ En + O(ε) (4. If the number of eigenvalues N is ﬁnite we set Ej = inf σess (A) for j > N and choose ϕj orthonormal such that (A − Ej )ϕj ≤ ε. ψ ∈ span{ψ1 . Then by Theorem 2. Thus the optimal choice ψ1 = ϕ1 will give the maximal value E2 . one can optimize them and obtain quite good upper bounds for the ﬁrst eigenvalue. n ψ. . But is there also something one can say about the next eigenvalues? Suppose we know the ﬁrst eigenfunction ϕ1 . More precisely. (4..34) In fact. . there will still be a component in the direction of ϕ1 left and hence the inﬁmum of the expectations will be lower than E2 . one often has some idea how the eigenfunctions might approximately look like. . where Ej ≤ Ej+1 are the eigenvalues (counted according to their multiplicity). . Aϕ1 = E1 . . However.32) If we add some free parameters to ψ1 ..
λ2 )) ≥ k. Problem 4.9 (MinMax). explaining the name. dim Ran PA ((−∞. are linearly independent elements of a H. ∞)) ≥ k.4. (4. (Hint A − An ≤ ε is equivalent to A − ε ≤ A ≤ A + ε. λ)) ≥ k. If (A − λ2 − λ1 λ2 + λ1 )ψ < ψ 2 2 k j=1 cj ψj .ψn−1 ) Clearly the same result holds if D(A) is replaced by the quadratic form domain Q(A) in the deﬁnition of U . Let λ1 < λ2 ..2. ψ. Aψ ≥ En − O(ε)... ψj ∈ Q(A). Let A be selfadjoint and let E1 ≤ E2 ≤ E3 · · · be the eigenvalues of A below the essential spectrum respectively the inﬁmum of the essential spectrum once there are no more eigenvalues left. Aψ < λ ψ for any nonzero linear combination ψ = 2 (4. Let λ ∈ R. Similarly. Suppose A. Applications of the spectral theorem (ii) We have inf ψ∈U (ϕ1 . Suppose A and B are selfadjoint operators with A ≥ B (i.. Suppose A is a selfadjoint operator and ψj . In addition. (ii).41) dim Ran PA ((λ1 . 1 ≤ j ≤ k.ψn−1 ψ∈U (ψ1 .. we show that the dimension of the range of PA (Ω) can be estimated if we have some functions which lie approximately in this space. (i). then En (A) ≥ En (B).ϕn−1 ) ψ..37) ψ1 . the sup and inf are in fact max and min. Aψ > λ ψ 2 implies dim Ran PA ((λ. (4. as long as En is an eigenvalue..104 4. Corollary 4. Aψ . A − B ≥ 0).. An are bounded and An → A. If ψ.38) then (4. Since ε can be chosen arbitrarily small we have proven Theorem 4..) 4. Then En = sup inf ψ.11. (4. set ψ = ϕn ..e.. .10. Theorem 4. Then Ek (An ) → Ek (A).36) In fact. Estimating eigenspaces Next. ψj ∈ D(A)..40) for any nonzero linear combination ψ = then (4.39) k j=1 cj ψj .
Then we see that for any nonzero linear combination ψ ψ. (4. . (4. which will again be denoted by P (A1 . . Tensor products of operators Suppose Aj .45) n and extend this deﬁnition to obtain a linear operator on the set D = span{ψ1 ⊗ · · · ⊗ ψn  ψj ∈ D(AN )}. that is. 1 ≤ j ≤ n. j (4. . We claim dim PA ((−∞. λ))M = {0}. An ) as above. (i). ∞). Using the same notation as before we need to show KerPA ((λ1 . Aψ = R η dµψ (η) = [λ. . Theorem 4. . are selfadjoint operators on Hj and let P (λ1 . . . λ))ψ = 0. . σ(An )). . . . . . .12. . . If PA ((λ1 . An )) = P (σ(A1 ). ψ = 0. . . . Suppose PA ((−∞. . 1 ≤ j ≤ n.5.43) 4 2 4 Ω where Ω = (−∞. (A − λ2 + λ 1 λ2 + λ1 2 λ2 + λ 1 )ψ 2 = (x − ) dµψ (x) = x2 dµψ (x + ) 2 2 2 R Ω (λ2 − λ1 )2 λ 2 + λ1 (λ2 − λ1 )2 ≥ dµψ (x + )= ψ 2. n n 1 1 ψj ∈ D(Aj j ). j (4. Suppose Aj . ψj ∈ D(AN ). . σ(P (A1 .47) . An ) on D is symmetric and we can consider its closure. then.46) Moreover.42) ≥ λ [λ. But this is a contradiction as before. .∞) dµψ (η) = λ ψ 2 . −(λ2 −λ1 )/2]∪[(λ2 −λ1 )/2. λ2 ))M = {0}. . . . . Then P (A1 .4. λn ) be a realvalued polynomial and deﬁne P (A1 . ψ = 0. 4. are selfadjoint operators on Hj .38). . For this it suﬃces to show KerPA ((−∞. . This contradicts our assumption (4. . For every monomial λn1 · · · λnn we can deﬁne n 1 (An1 ⊗ · · · ⊗ Ann )ψ1 ⊗ · · · ⊗ ψn = (An1 ψ1 ) ⊗ · · · ⊗ (Ann ψn ). An )ψ1 ⊗ · · · ⊗ ψn . . .5. then the operator P (A1 . . (4. if P is realvalued.∞) η dµψ (η) (4. λn ) of degree N we can deﬁne P (A1 . An ) is selfadjoint and its spectrum is the closure of the range of P on the product of the spectra of the Aj . Tensor products of operators 105 Proof. λ))M = dim M = k. . An ). λ2 ))ψ = 0. . . (ii). Let M = span{ψj } ⊆ H.44) Hence for every polynomial P (λ1 . .
Now let λ = P (λ1 . . Hence P is c the maximally deﬁned multiplication operator by P (λ1 . dµ1 × · · · × dµn ). which is selfadjoint. . . . λj with respect to µj and hence (P −λ)−1 exists and is bounded. .106 4. Then.k ∈ D(AN ) with (Aj − λj )ψj. . .k and hence λ ∈ σ(P ). where ψk = ψ1. . The two main cases of interest are A1 ⊗ A2 . σ(An )). in which case σ(A1 ⊗ I + I ⊗ A2 ) = σ(A1 ) + σ(A2 ) = {λ1 + λ2 λj ∈ σ(Aj )}. dµ1 × · · · × dµn ). dµj ) it follows that the c domain of the closure of P contains L2 (Rn . Applications of the spectral theorem Proof. . in which case σ(A1 ⊗ A2 ) = σ(A1 )σ(A2 ) = {λ1 λ2 λj ∈ σ(Aj )}. . . λn ). .49) (4. An ) is hence multiplication by P (λ1 . dµj ) and P (A1 . . . if λ ∈ P (σ(A1 ). and A1 ⊗ I + I ⊗ A2 . then P (λ1 . By the spectral theorem it is no restriction to assume that Aj is multiplication by λj on L2 (R. . Conversely. . Then there exists Weyl sequences ψj.e.48) . λn ) − λ ≥ ε for a.k → 0 as k → ∞. . Since D contains the set of all functions ψ1 (λ1 ) · · · ψn (λn ) for which ψj ∈ L2 (R. (P − j λ)ψk → 0. . . .k ⊗ · · · ⊗ ψ1. . (4. . . λn ) on L2 (Rn . λn ) with λj ∈ σ(Aj ). . that is λ ∈ ρ(P ).
12) in the Hilbert space H. U (t) is a strongly continuous oneparameter unitary group.3) Our hope is that this formula also applies in the general case and that we can reconstruct a oneparameter unitary group U (t) from its generator A (compare (2.1) 5.1. 107 .2) A detailed investigation of this formula will be our ﬁrst task.e. We ﬁrst investigate the family of operators exp(−itA). our system is not hyperbolic (i. (i).11)) via U (t) = exp(−itA). the solution of the Schr¨dinger equation o i is given by ψ(t) = exp(−itH)ψ(0).. the spectrum is not away from the real axis) and hence simple results like.1. all solutions tend to the equilibrium position cannot be expected. Let A be selfadjoint and let U (t) = exp(−itA). Theorem 5. in the ﬁnite dimensional case the dynamics is understood once the eigenvalues are known and the same is true in our case once we know the spectrum. The time evolution and Stone’s theorem In this section we want to have a look at the initial value problem associated with the Schr¨dinger equation (2. (5. like any Hamiltonian system from classical mechanics. the solution is given by ψ(t) = e−itA ψ(0). Note that. (5.Chapter 5 Quantum dynamics As in the ﬁnite dimensional case. Moreover. d ψ(t) = Hψ(t) dt (5. If H is oneo dimensional (and hence A is a real number).
This is ensured by Stone’s theorem. Aψ (5. This corresponds to conservation of energy.6) ϕ. Similarly. This settles (ii). On the other hand. Then ˜ A is a symmetric extension of A since we have i i ˜ ˜ (U (−t) − 1)ϕ. First of all observe that weak continuity together with Lemma 1.3) is indeed the solution to the initial value problem of the Schr¨dinger equation.4) by the dominated convergence theorem. there should be a one to one correspondence between the unitary group and its generator. Then its generator A is selfadjoint and U (t) = exp(−itA). To see (iii) replace ψ → U (s)ψ in (ii). . Moreover.2. U (t)D(A) = D(A) and AU (t) = U (t)A. Moreover.108 5. Proof. o U (t)ψ.1 and the corresponding statements for the function exp(−itλ). For our original problem this implies that formula (5.5) t ˜ since eitλ − 1 ≤ tλ. To prove strong continuity observe that t→t0 lim e−itA ψ − e−it0 A ψ 2 = = t→t0 lim e−itλ − e−it0 λ 2 dµψ (λ) R R t→t0 lim e−itλ − e−it0 λ 2 dµψ (λ) = 0 (5. Aψ = lim ϕ. Quantum dynamics (ii). Proof. Let U (t) be a weakly continuous oneparameter unitary group. (U (t) − 1)ψ = lim t→0 −t t→0 t ˜ and hence A = A by Corollary 2. AU (t)ψ = U (t)ψ.11). The limit limt→0 1 (U (t)ψ − ψ) exists if and only if ψ ∈ D(A) in t which case limt→0 1 (U (t)ψ − ψ) = −iAψ. if ψ ∈ D(A) we obtain t→0 lim 1 −itA (e ψ − ψ) + iAψ t 2 = lim t→0 R 1  (e−itλ − 1) + iλ2 dµψ (λ) = 0 (5. Now let A be the generator deﬁned as in (2. ψ = Aϕ. t (iii). U (t)Aψ = ψ.7) shows that the expectations of A are time independent. ψ (5.11 (iv) shows that U (t) is in fact strongly continuous.2 (Stone). the generator of the time evolution of a quantum mechanical system should always be a selfadjoint operator since it corresponds to an observable (energy). Theorem 5. The group property (i) follows directly from Theorem 3.
ψ = 0 implying ϕ = 0 since D(A) is dense. U (t)ψ dt = ϕ. . Then ψ(t + s) − ψ(t) = iAψ(t) s→0 s lim (5. Corollary 5.8) (the integral is deﬁned as in Section 4. Then A is essentially selfadjoint on D. Proof.10) = −iz ϕ. U (t)ψ and hence ϕ. So A is essentially selfadjoint and we can introduce V (t) = exp(−itA).3. iAψ(t) = 0. As in the above proof it follows ϕ.1) implying limτ →0 τ −1 ψτ = ψ. Pick ψ ∈ H and set τ ψτ = 0 U (t)ψdt (5.6 it suﬃces to prove Ker(A∗ − z ∗ ) = {0} for z ∈ C\R. we must have ϕ. As an immediate consequence of the proof we also note the following useful criterion. −iAU (t)ψ = −i A∗ ϕ. ψ = 0 for any ϕ ∈ Ker(A∗ − z ∗ ) and ψ ∈ D.11) d and hence dt ψ(t) 2 = 2Re ψ(t).9) as t → 0 shows ψτ ∈ D(A). Since the left hand side is bounded for all t ∈ R and the exponential on the right hand side is not. U (t)ψ (5.5. We are done if we can show U (t) = V (t). Moreover. By Lemma 2. As in the proof of the previous theorem. The time evolution and Stone’s theorem 109 Next we show that A is densely deﬁned. Next. then for each ψ ∈ D(A) we have d ϕ. we can show that A is symmetric and that U (t)D(A) = D(A). ψ . let us prove that A is essentially selfadjoint. U (t)ψ = exp(−izt) ϕ. Furthermore. 1 τ 1 t+τ 1 U (s)ψds − U (s)ψds (U (t)ψτ − ψτ ) = t t t t 0 1 τ +t 1 t = U (s)ψds − U (s)ψds t τ t 0 t 1 1 t = U (τ ) U (s)ψds − U (s)ψds → U (τ )ψ − ψ t t 0 0 (5. since D(A) is dense we have U (t) = V (t) by continuity. Suppose A∗ ϕ = z ∗ ϕ. Let ψ ∈ D(A) and abbreviate ψ(t) = (U (t) − V (t))ψ.1. Suppose D ⊆ D(A) is dense and invariant under U (t). Since ψ(0) = 0 we have ψ(t) = 0 and hence U (t) and V (t) coincide on D(A).
(5. Problem 5. Quantum dynamics Note that by Lemma 4.2.1. Thus our question is answered by Wiener’s theorem. Let µ be a ﬁnite complex Borel measure on R and let µ(t) = ˆ R e−itλ dµ(λ) (5. (5. how does  ϕ.14) is the Fourier transform of the measure µϕ. o To understand the dynamics of such a ﬁrst order diﬀerential equation. let us discuss why the decomposition of the spectrum introduced in Section 3. 2π) and consider the one parameter unitary group given by U (t)f (x) = f (x − t mod 2π). Then the Ces`ro time average of µ(t) has the a ˆ following limit 1 T →∞ T lim T ˆ(t)2 dt = µ 0 λ∈R µ({λ})2 . . the Schr¨dinger equation.13) behave as t → ∞? By the spectral theorem. for a physicist. Clearly. U (t) = e−itA . We have seen that the time evolution is generated by a selfadjoint operator.3 is of physical relevance. A ﬁrst question one might rise is.ψ . Hence  ϕ.4 (Wiener). Let H = L2 (0.110 5. ψ 2 can be viewed as the part of ψ which is in the state ϕ. e−isB ] = 0 if and only if the generators commute. and is given by a linear ﬁrst order diﬀerential equation. U (t)ψ = ˆ R e−itλ dµϕ. ψ ϕ is the projection of ψ onto the (onedimensional) subspace spanned by ϕ.15) be its Fourier transform. The RAGE theorem Now.16) where the sum on the right hand side is ﬁnite. Some general tools for this endeavor will be provided in the following sections. What is the generator of U ? (5. U (t)ψ 2 . µϕ. The vector ϕ.12) 5. one of the goals must be to understand the time evolution of a quantum mechanical system.ψ (t) = ϕ.8 two strongly continuous oneparameter groups commute [e−itA .ψ (λ) (5. Theorem 5. Let ϕ = ψ = 1. one must understand the spectrum of the generator. the Hamiltonian.
f (A)ψ implying dµϕ.6 below). If {ψj }n j=1 is an orthonormal basis for Ran(K) we have n n Kψ = j=1 ψj . pp}. and Hpp are invariant with respect to time evolution since P xx U (t) = χMxx (A) exp(−itA) = exp(−itA)χMxx (A) = U (t)P xx . (5. But ﬁrst we need a few deﬁnitions.18) To apply this result to our situation. If ψ ∈ Hac .ψ . xx ∈ {ac. or pp. observe that the subspaces Hac . In ˆ fact.2. if ψ ∈ Hxx we have P xx ψ = ψ which shows ϕ. The RAGE theorem 111 Proof. U (t)ψ tends a to zero. so is µϕ.ψ (t) is continuous and tends to zero as t → ∞. the limit of the above expression is given by χ{0} (x − y)dµ(x)dµ∗ (y) = R R R µ({y})dµ∗ (y) = y∈R µ({y})2 . Kψ ψj = j=1 ϕj . Moreover. f (A)ψ = ϕ. P xx f (A)ψ = P xx ϕ. Now we want to draw some additional consequences from Wiener’s theorem. This will eventually yield a dynamical characterization of the continuous and pure point spectrum due to Ruelle. An operator K ∈ L(H) is called ﬁnite rank if its range is ﬁnite dimensional. Amrein. Thus if µψ is ac.19) . In other words. by the dominated convergence theorem.ψ = dµP xx ϕ. That is. the average of the probability of ﬁnding the system in any prescribed state tends to zero if we start in the continuous subspace Hc of A. (5. then the Ces`ro mean of ϕ.ψ for every ϕ ∈ H. and Enß. The dimension n = dim Ran(K) is called the rank of K. sc. Hsc . Gorgescu. this follows from the RiemannLebesgue lemma (see Lemma 7.17) The function in parentheses is bounded by one and converges pointwise to χ{0} (x − y) as T → ∞.5.ψ is absolutely continuous with respect to Lebesgue measure and thus µϕ. if ψ ∈ Hc = Hac ⊕Hsc . ψ ψ j . (5. Thus. sc. then dµϕ. By Fubini we have 1 T T ˆ(t)2 dt = µ 0 1 T T 0 R R e−i(x−y)t dµ(x)dµ∗ (y)dt 1 T T 0 = R R e−i(x−y)t dt dµ(x)dµ∗ (y).
the claim can be reduced to the previous situation. ϕ2 ). The elements ϕj are linearly independent since Ran(K) = Ker(K ∗ )⊥ .e. K is bounded.2) Lemma 5. Theorem 5. It is straightforward to verify (Problem 5. The set of all compact operators C(H) is a closed ∗ideal in L(H). K = ϕ1 . In particular.19).6.112 5. in addition. the claim follows from Wiener’s theorem respectively the RiemannLebesgue lemma. there is a sequence Kn of ﬁnite rank operators such that K − Kn ≤ 1/n and hence 1 Ke−itA ψ ≤ Kn e−itA ψ + ψ . ψ ϕj .20) The closure of the set of all ﬁnite rank operators in L(H) is called the set of compact operators C(H). Then 1 T lim Ke−itA P c ψ 2 dt = 0 and lim Ke−itA P ac ψ = 0 t→∞ T →∞ T 0 (5.21) for one z ∈ ρ(A). Then. If K is compact. By the ﬁrst resolvent identity this then follows for all z ∈ ρ(A). if K is also bounded. Now let us return to our original problem. If. Hence it holds for any ﬁnite rank operator K. (5. (5. There is also a weaker version of compactness which is useful for us. Hence every ﬁnite rank operator is of the form (5. Let ψ ∈ Hc respectively ψ ∈ Hac and drop the projectors. Let A be selfadjoint and suppose K is relatively compact. Since K(A + i)−1 is compact by assumption. where ϕ ∈ Hc if and only if ψ ∈ Hc (since Hc reduces A). In particular we have D(A) ⊆ D(K).24) n . the adjoint of K is also ﬁnite rank and given by n K ∗ψ = j=1 ψ j . we can ﬁnd a sequence ψn ∈ D(A) such that ψ − ψn ≤ 1/n and hence 1 Ke−itA ψ ≤ Ke−itA ψn + K . Proof. then the result holds for any ψ ∈ H. Quantum dynamics where ϕj = K ∗ ψj . if K is a rank one operator (i.5. The operator K is called relatively compact with respect to A if KRA (z) ∈ C(H) (5.22) for every ψ ∈ D(A).. In addition. If ψ ∈ D(A) we can set ψ = (A − i)−1 ϕ.23) n Thus the claim holds for any compact operator K. . (5.
then 1 T T T 1/2 Kn ψ(t) dt ≤ 0 1 T Kn ψ(t) 2 dt 0 →0 (5. (5.2. Then Hc = {ψ ∈ H lim lim n→∞ t≥0 1 n→∞ T →∞ T T 0 Kn e−itA ψ dt = 0}. Theorem 5. we can write ψ pp (t) = j αj (t)ψj . then this part converges uniformly to the desired limit by strong convergence of Kn . if ψ ∈ Hpp we can write ψ = ψ c + ψ pp and by our previous estimate it suﬃces to show that (I − Kn )ψ c (t) does not tend to 0 as n → ∞.27) By the spectral theorem. if ψ ∈ Hc we can write ψ = ψ c + ψ pp . If it would.28) . 1 T →∞ T lim ψ c (t) 2 T (I − Kn )ψ c (t) 2 dt 0 − lim 1 T →∞ T T Kn ψ c (t) 2 dt = ψ c (t) 2 . we even claim n→∞ t≥0 lim sup Kn ψ pp (t) − ψ pp (t) = 0. Now let us turn to the second equation. In summary. regularity properties of spectral measures are related to the long time behavior of the corresponding quantum mechanical system. and hence the error can be made arbitrarily small by choosing N large. With the help of this result we can now prove an abstract version of the RAGE theorem. Let ψ ∈ Hc .7 (RAGE).13 we have Kn ≤ M . If ψ ∈ Hpp the claim follows by (5. In fact.25) Hpp = {ψ ∈ H lim sup (I − Kn )e−itA ψ = 0}. Truncate this expansion after N terms. Conversely. The RAGE theorem 113 concluding the proof. by Lemma 1. (5. Conversely. Let A be selfadjoint. we would have 0 = ≥ a contradiction. By our previous estimate it suﬃces to show Kn ψ pp (t) ≥ ε > 0 for n large.27). Abbreviate ψ(t) = exp(−itA)ψ.26) by CauchySchwarz and the previous theorem. Suppose Kn ∈ L(H) is a sequence of relatively compact operators which converges strongly to the identity. 0 (5. We begin with the ﬁrst equation. Moreover. Proof.5. where the ψj are orthonormal eigenfunctions and αj (t) = exp(−itλj )αj .
6. a more detailed investigation of this topic is beyond the scope of this manuscript. we can interchange the limit with the summation and the claim follows. The strong limits are called asymptotic observables if they exist.32) by Theorem 5.31) If K is in addition bounded. (5. For a survey containing several recent results see [9]. Proof. Write ψ = ψ c + ψ pp .9. Quantum dynamics However. we see that the operator in parenthesis tends to PA ({λj }) strongly as T → ∞.34) n→∞ T →∞ T 0 . K(t)ψ . use the same trick as before and replace K by KRA (z). As in the proof of the previous theorem we can write ψ pp = j αj ψj and hence αj j 1 T T K(t)ψj dt = 0 j αj 1 T T eit(A−λj ) dt Kψj . Under the same assumptions as in the RAGE theorem we have 1 T itA lim lim e Kn e−itA ψdt = P pp ψ (5. The main interest is the behavior of K(t) for large t. We will assume that K is bounded.29) and note ψ(t). Then 1 T →∞ T lim T 0 K(t)ψ c dt ≤ lim 1 T →∞ T T K(t)ψ c dt = 0 0 (5. the result holds for any ψ ∈ H. Kψ(t) = ψ. Corollary 5.30) This point of view is often referred to as Heisenberg picture in the physics literature. Since this operator is also bounded by 1 for all T . Theorem 5.33) As in the proof of Wiener’s theorem. It is often convenient to treat the observables as time dependent rather than the states. ψ(t) = e−itA ψ. We also note the following corollary. 0 (5. Then 1 T →∞ T lim T 0 eitA Ke−itA ψdt = λ∈σp (A) PA ({λ})KPA ({λ})ψ. If K is unbounded we will assume D(A) ⊆ D(K) such that the above equations make sense at least for ψ ∈ D(A). ψ ∈ D(A). To obtain the general result. (5. Suppose A is selfadjoint and K is relatively compact.114 5.8. We set K(t) = eitA Ke−itA (5.
So limτ →0 Fτ (s) = 0 at least pointwise.36) Proof.37) j where τ = t n. However.39) τ Now for ψ ∈ D(A + B) = D(A) ∩ D(B) we have 1 iτ A iτ B (e e − eiτ (A+B) )ψ → iAψ + iBψ − i(A + B)ψ = 0 (5.3. Since A and B will not commute in general.3. we cannot obtain eit(A+B) from eitA eitB . A. Suppose. Fτ (s) = Pointwise convergence implies 1 iτ A iτ B (e e − eiτ (A+B) )ψ ≤ C(ψ) (5. (5.35) Problem 5.10 (Trotter product formula). but we need this uniformly with respect to s ∈ [−t. and hence (eiτ A eiτ B )n − eit(A+B) ψ ≤ t max Fτ (s). we can invoke the uniform boundedness ψ 2 Γ(A+B) = ψ principle to obtain 1 iτ A iτ B (e e − eiτ (A+B) )ψ ≤ C ψ Γ(A+B) . The Trotter product formula 115 respectively 1 n→∞ T →∞ T lim lim T 0 e−itA (I − Kn )e−itA ψdt = P c ψ.38) 1 iτ A iτ B (e e − eiτ (A+B) )eis(A+B) ψ . B. (5. since D(A + B) is a Hilbert space when equipped with the graph norm 2 + (A + B)ψ 2 . Problem 5. (5. Then eit(A+B) = slim ei n A ei n B n→∞ t t n . t]. we at least have: Theorem 5. s≤t (5.42) τ where .(5.40) τ as τ → 0. (5. where eitA and eitB can be computed explicitly.3.5. Prove Corollary 5. Prove Lemma 5. The Trotter product formula In many situations the operator is of the form A + B. 5.5.2.9. First of all note that we have eiτ A eiτ B n−1 n − eit(A+B) n−1−j = j=0 eiτ A eiτ B eiτ A eiτ B − eiτ (A+B) eiτ (A+B) .41) τ and. and A + B are selfadjoint.
116
5. Quantum dynamics
Now Fτ (s) − Fτ (r) ≤ 1 iτ A iτ B (e e − eiτ (A+B) )(eis(A+B) − eir(A+B) )ψ τ ≤ C (eis(A+B) − eir(A+B) )ψ Γ(A+B) . (5.43)
shows that Fτ (.) is uniformly continuous and the claim follows by a standard ε 2 argument. If the operators are semibounded from below the same proof shows Theorem 5.11 (Trotter product formula). Suppose, A, B, and A + B are selfadjoint and semibounded from below. Then e−t(A+B) = slim e− n A e− n B
n→∞
t t
n
,
t ≥ 0.
(5.44)
Problem 5.4. Proof Theorem 5.11.
Chapter 6
Perturbation theory for selfadjoint operators
The Hamiltonian of a quantum mechanical system is usually the sum of the kinetic energy H0 (free Schr¨dinger operator) plus an operator V coro responding to the potential energy. Since H0 is easy to investigate, one usually tries to consider V as a perturbation of H0 . This will only work if V is small with respect to H0 . Hence we study such perturbations of selfadjoint operators next.
6.1. Relatively bounded operators and the Kato–Rellich theorem
An operator B is called A bounded or relatively bounded with respect to A if D(A) ⊆ D(B) and if there are constants a, b ≥ 0 such that Bψ ≤ a Aψ + b ψ , ψ ∈ D(A). (6.1)
The inﬁmum of all constants a for which a corresponding b exists such that (6.1) holds is called the Abound of B. The triangle inequality implies Lemma 6.1. Suppose Bj , j = 1, 2, are A bounded with respective Abounds ai , i = 1, 2. Then α1 B1 + α2 B2 is also A bounded with Abound less than α1 a1 + α2 a2 . In particular, the set of all A bounded operators forms a linear space. There are also the following equivalent characterizations: 117
118
6. Perturbation theory for selfadjoint operators
Lemma 6.2. Suppose A is closed and B is closable. Then the following are equivalent: (i) B is A bounded. (ii) D(A) ⊆ D(B). (iii) BRA (z) is bounded for one (and hence for all) z ∈ ρ(A). Moreover, the Abound of B is no larger then inf z∈ρ(A) BRA (z) . Proof. (i) ⇒ (ii) is true by deﬁnition. (ii) ⇒ (iii) since BRA (z) is a closed (Problem 2.6) operator deﬁned on all of H and hence bounded by the closed graph theorem (Theorem 2.7). To see (iii) ⇒ (i) let ψ ∈ D(A), then Bψ = BRA (z)(A − z)ψ ≤ a (A − z)ψ ≤ a Aψ + (az) ψ , (6.2) where a = BRA (z) . Finally, note that if BRA (z) is bounded for one z ∈ ρ(A), it is bounded for all z ∈ ρ(A) by the ﬁrst resolvent formula. We are mainly interested in the situation where A is selfadjoint and B is symmetric. Hence we will restrict our attention to this case. Lemma 6.3. Suppose A is selfadjoint and B relatively bounded. The Abound of B is given by
λ→∞
lim BRA (±iλ) .
(6.3)
If A is bounded from below, we can also replace ±iλ by −λ. Proof. Let ϕ = RA (±iλ)ψ, λ > 0, and let a∞ be the Abound of B. If B is A bounded, then (use the spectral theorem to estimate the norms) BRA (±iλ)ψ ≤ a ARA (±iλ)ψ + b RA (±iλ)ψ ≤ (a + b ) ψ . λ (6.4)
Hence lim supλ BRA (±iλ) ≤ a∞ which, together with a∞ ≤ inf λ BRA (±iλ) from the previous lemma, proves the claim. The case where A is bounded from below is similar. Now we will show the basic perturbation result due to Kato and Rellich. Theorem 6.4 (Kato–Rellich). Suppose A is (essentially) selfadjoint and B is symmetric with Abound less then one. Then A + B, D(A + B) = D(A), is (essentially) selfadjoint. If A is essentially selfadjoint we have D(A) ⊆ D(B) and A + B = A + B. If A is bounded from below by γ, then A + B is bounded from below by min(γ, b/(a − 1)).
6.2. More on compact operators
119
Proof. Since D(A) ⊆ D(B) and D(A) ⊆ D(A + B) by (6.1) we can assume that A is closed (i.e., selfadjoint). It suﬃces to show that Ran(A+B ±iλ) = H. By the above lemma we can ﬁnd a λ > 0 such that BRA (±iλ) < 1. Hence −1 ∈ ρ(BRA (±iλ)) and thus I + BRA (±iλ) is onto. Thus (A + B ± iλ) = (I + BRA (±iλ))(A ± iλ) is onto and the proof of the ﬁrst part is complete. If A is bounded from below we can replace ±iλ by −λ and the above equation shows that RA+B exists for λ suﬃciently large. By the proof of the previous lemma we can choose −λ < min(γ, b/(a − 1)). Finally, let us show that there is also a connection between the resolvents. Lemma 6.5. Suppose A and B are closed and D(A) ⊆ D(B). Then we have the second resolvent formula RA+B (z) − RA (z) = −RA (z)BRA+B (z) = −RA+B (z)BRA (z) for z ∈ ρ(A) ∩ ρ(A + B). Proof. We compute RA+B (z) + RA (z)BRA+B (z) = RA (z)(A + B − z)RA+B (z) = RA (z). (6.7) The second identity is similar. Problem 6.1. Compute the resolvent of A + α ψ, . ψ. (Hint: Show α (I + α ϕ, . ψ)−1 = I − ϕ, . ψ 1 + α ϕ, ψ and use the second resolvent formula.) (6.6) (6.5)
(6.8)
6.2. More on compact operators
Recall from Section 5.2 that we have introduced the set of compact operators C(H) as the closure of the set of all ﬁnite rank operators in L(H). Before we can proceed, we need to establish some further results for such operators. We begin by investigating the spectrum of selfadjoint compact operators and show that the spectral theorem takes a particular simple form in this case. We introduce some notation ﬁrst. The discrete spectrum σd (A) is the set of all eigenvalues which are discrete points of the spectrum and whose corresponding eigenspace is ﬁnite dimensional. The complement of the discrete spectrum is called the essential spectrum σess (A) = σ(A)\σd (A). If A is selfadjoint we might equivalently set σd (A) = {λ ∈ σp (A)rank(PA ((λ − ε, λ + ε))) < ∞ for some ε > 0}. (6.9)
120
6. Perturbation theory for selfadjoint operators
respectively σess (A) = {λ ∈ Rrank(PA ((λ − ε, λ + ε))) = ∞ for all ε > 0}. (6.10)
Theorem 6.6 (Spectral theorem for compact operators). Suppose the operator K is selfadjoint and compact. Then the spectrum of K consists of an at most countable number of eigenvalues which can only cluster at 0. Moreover, the eigenspace to each nonzero eigenvalue is ﬁnite dimensional. In other words, σess (K) ⊆ {0}, where equality holds if and only if H is inﬁnite dimensional. In addition, we have K=
λ∈σ(K)
(6.11)
λPK ({λ}).
(6.12)
Proof. It suﬃces to show rank(PK ((λ − ε, λ + ε))) < ∞ for 0 < ε < λ. Let Kn be a sequence of ﬁnite rank operators such that K −Kn ≤ 1/n. If RanPK ((λ−ε, λ+ε)) is inﬁnite dimensional we can ﬁnd a vector ψn in this range such that ψn = 1 and Kn ψn = 0. But this yields a contradiction 1 ≥  ψn , (K − Kn )ψn  =  ψn , Kψn  ≥ λ − ε > 0 n by (4.2). As a consequence we obtain the canonical form of a general compact operator. Theorem 6.7 (Canonical form of compact operators). Let K be compact. ˆ There exists orthonormal sets {φj }, {φj } and positive numbers sj = sj (K) such that ˆ ˆ K= sj φj , . φj , K∗ = sj φj , . φj . (6.14)
j j
(6.13)
ˆ ˆ ˆ Note Kφj = sj φj and K ∗ φj = sj φj , and hence K ∗ Kφj = s2 φj and KK ∗ φj = j ˆ s2 φj . j The numbers sj (K)2 > 0 are the nonzero eigenvalues of KK ∗ respectively K ∗ K (counted with multiplicity) and sj (K) = sj (K ∗ ) = sj are called singular values of K. There are either ﬁnitely many singular values (if K is ﬁnite rank) or they converge to zero. Proof. By Lemma 5.5 K ∗ K is compact and hence Theorem 6.6 applies. Let {φj } be an orthonormal basis of eigenvectors for PK ∗ K ((0, ∞))H and let
note that we have K = max sj (K).12. let me remark that there are a number of other equivalent deﬁnitions for compact operators. ψ 2 An φj 2 ≤ j=1 ˆ sj An φj 2 → 0.17) Finally. j (6. .2. If ψn is bounded it has a weakly convergent subsequence by Lemma 1. (i) ⇔ (i’). (iv) ψn bounded implies that Kψn has a (norm) convergent subsequence. since K ψ 2 = ψ. for any ψ ∈ H we can write j ψ= j ˜ φj . (iii) ⇒ (iv).16) ˆ ˜ ˜ ˜ ˆ ˆ where φj = s−1 Kφj . ψ φj . Then (by (6. This is immediate from Theorem 6. φk = j (sj sk )−1 Kφj . (i) ⇒ (ii). Proof. Since An ≤ M it suﬃces to consider the case where K is ﬁnite rank. Kφk = (sj sk )−1 K ∗ Kφj . replace ψn → ψn − ψ and assume ψ = 0. Again. More on compact operators 121 s2 be the eigenvalue corresponding to φj . Moreover.8. 2 ˆ If K is selfadjoint then φj = σj φj . By φj .7. Choose An = ψn . ϕ = 1. (iii) ψn ψ weakly implies Kψn → Kψ in norm. φk we see that k ˆ {φj } are orthonormal and the formula for K ∗ follows by taking the adjoint of the formula for K (Problem 6. (6. (6. φk = sj s−1 φj .18) (ii) ⇒ (iii). Then. then Kψn 2 = An K ∗ → 0. (i’) K ∗ is compact. (ii) An ∈ L(H) and An → A strongly implies An K → AK. K ∗ K ψ = 0.15) ˜ with ψ ∈ Ker(K ∗ K) = Ker(K). Now apply (iii) to this subsequence. Translating An → An −A it is no restriction to assume A = 0. For K ∈ L(H) the following statements are equivalent: (i) K is compact.6.2). . ϕ. Then Kψ = j ˆ sj φj . σj = 1 are the eigenvectors of K and σj sj are the corresponding eigenvalues.14)) N N s An K 2 = sup ψ =1 j=1 ˆ sj  φj . ψ φj + ψ (6. Lemma 6.
By j=n assumption. converges to 0. Problem 6. where δ1 = (1. . 0. Using CauchySchwarz.3. Then.122 6. pick an orthonormal basis ϕj (x) for L2 (M. . . Kϕj . 2 Kψ(x)2 dµ(x) = M M M K(x. y)ψ(y)dµ(y) dµ(x) K(x. ).19) Then γn = K − Kn = sup ψ∈span{ϕj }∞ . δ1 . dµ). Then slim An = 0 but KAn = 1. 6. (6. Such an operator is called Hilbert– Schmidt operator. Hence ε must be 0 and we are done. ψ ∈ L2 (M. Deduce the formula for K ∗ from the one for K in (6. ψ =1 j=n Kψ (6. (6. dµ). Perturbation theory for selfadjoint operators (iv) ⇒ (i). y)2 dµ(y) ψ(y)2 dµ(y) dµ(x) M ≤ M M = M M K(x. y) ∈ L2 (M × M. Let ϕj be an orthonormal basis and set n Kn = j=1 ϕj . . y)2 dµ(y) dµ(x) M ψ(y)2 dµ(y) (6. ϕi (x)ϕj (y) is an orthonormal basis for . Example. Hilbert–Schmidt and trace class operators Among the compact operators two special classes or of particular importance. Next. The ﬁrst one are integral operators Kψ(x) = M K(x. Moreover. since ψn converges weakly to 0. y)ψ(y)dµ(y). we can ﬁnd a sequence of unit vectors ψn ∈ span{ϕj }∞ for which Kψn ≥ ε. An the operator which shifts each sequence n places to the left. The last condition explains the name compact.20) is a decreasing sequence tending to a limit ε ≥ 0. . Kψn has a convergent subsequence which.22) we see that K is bounded. Let H = 2 (N).9.2. Without the requirement that An is normal the claim is wrong as the following example shows. Moreover. note that you cannot replace An K → AK by KAn → KA unless you additionally require An to be normal (then this follows by taking adjoints — recall that only for normal operators taking adjoints is continuous with respect to strong convergence). dµ ⊕ dµ). by Lemma 1.14).21) where K(x. and K = δ1 .
y)2 dµ(y) dµ(x) < ∞. ci. (6.28) Then Kn has the kernel Kn (x. . Hence we will call a compact operator Hilbert–Schmidt if its singular values satisfy sj (K)2 < ∞. j (6. y) = i.24) In particular.29) By our lemma this coincides with our previous deﬁnition if H = L2 (M.25) shows that K can be approximated by ﬁnite rank operators (take ﬁnitely many terms in the sum) and is hence compact.26) in this case. Hilbert–Schmidt and trace class operators 123 L2 (M × M. dµ).23) where ci. Using (6.j ci. (6.j ϕ∗ .j = ϕi . If H = L2 (M.26) holds in this case. Kϕ∗ . Lemma 6. Kψ(x) = i. dµ ⊕ dµ) and K(x.14) we can also give a diﬀerent characterization of Hilbert– Schmidt operators. n ∗ˆ j=1 sj φj (y) φj (x) n (6.6.j ϕi (x)ϕj (y). y)2 dµ(x)dµ(y). so does the other and in particular. .j ci. (6.j 2 = i. dµ).j M M K(x.14): n Kn = j=1 ˆ sj φj . j (6.3. y)2 dµ(x)dµ(y) = M M j=1 sj (K)2 Now if one side converges. y) = Kn (x. If K is compact we can deﬁne approximating ﬁnite rank operators Kn by considering only ﬁnitely many terms in (6. then a compact operator K is Hilbert– Schmidt if and only if j sj (K)2 < ∞ and sj (K)2 = j M M K(x. ψ ϕi (x) j (6.27) and (6.9. Proof. φj .
(6. A compact operator K is Hilbert–Schmidt if and only if Kψn n 2 <∞ (6. (6.32) in this case. Let K be Hilbert–Schmidt and A bounded.j ˆ  K ∗ φj .j ˆ  φj . This approach can be generalized by deﬁning K plus corresponding spaces Jp (H) = {K ∈ C(H) K p p = j sj (K)p 1/p (6. Kψn 2 = n. (6. Proof.124 6.34) Proof. The set of Hilbert–Schmidt operators forms a ∗ideal in L(H) and KA 2 ≤ A K 2 respectively AK 2 ≤ A K 2.30) form a Hilbert space (isomorphic to L2 (M ×M. This follows from Kψn n 2 = n.35) For KA just consider adjoints.11.33) = n ˆ K ∗ φn 2 = j Corollary 6.10. ψn 2 sj (K)2 .37) . Perturbation theory for selfadjoint operators Since every Hilbert space is isomorphic to some L2 (M. Then AK is compact and AK 2 2 = n AKψn 2 ≤ A 2 n Kψn 2 = A 2 K 2 2. Note that K 2 = K ∗ 2 (since sj (K) = sj (K ∗ )). (6.36) < ∞}. There is another useful characterization for identifying Hilbert–Schmidt operators: Lemma 6. dµ) we see that the Hilbert–Schmidt operators together with the norm K 2 = j sj (K)2 1/2 (6.31) for some orthonormal set and Kψn n 2 = K 2 2 (6. dµ⊗dµ)).
Hilbert–Schmidt and trace class operators 125 which are known as Schatten pclass. φj 2 j 1/2 j ˆ sp  ψn ..40) holds. φj 2 j sp  ϕn ..40): Choose q such that p + 1 = 1 and q use H¨lder’s inequality to obtain (sj . φj 2 ≤ j j sp  ϕn . φj 2 j = j 1/2 ≤ sp j 1/2 sp . ψn p ≤ j sp  ϕn . Kϕj p j where the sup is taken over all orthonormal systems.43) ˆ Since equality is attained for ϕn = φn and ψn = φn equation (6. Now using CauchySchwarz we have  ψn . (6.6.38) = K ∗ p.. The hard part is to prove (6. φj 2 . (6.. 1 Proof.j sp  ϕn .j ˆ sp  ψn .42) Summing over n.41) ˆ Clearly the analogous equation holds for φj . (6. Kϕn p = j sj 1/2 ϕn . Kϕn p ≤ n n..39) Lemma 6.2/q ) o j sj  ϕn . {ϕj } ONS . ψn .12.. 1/p {ψj }. φj 2 j sp j j 1/2 j 1/2 n. φj sj 1/2 ˆ φj . Moreover. 1/q ≤ (6. Note that K ≤ K and that by sj (K) = sj (K ∗ ) we have K p p.2 )1/p . p are Banach spaces. a second appeal to CauchySchwarz and interchanging the order of summation ﬁnally gives  ψn . φj 2 j j 1/p j 1/p  ϕn .2 = (sp . The spaces Jp (H) together with the norm .3. φj 2 j 1/2 (6.40) K p = sup  ψj . j (6. .
there is a compact K with K − Kn → 0 and by Kn p ≤ C we have  ψj . Since Hilbert–Schmidt operators are easy to identify it is important to relate J1 (H) with J2 (H): Lemma 6. . The set of trace class operators forms a ∗ideal in L(H) and KA 1 ≤ A K 1 respectively AK 1 ≤ A K 1. Perturbation theory for selfadjoint operators Now the rest is straightforward.46) K2 ψ n 2 1/2 = K1 K2 2 (6. ( K ≤ K p ). K 2 ψ n  ≤ n 2 ∗ K 1 ϕn 2 n 1 ≤ K1 2 K2 2 (6.48) . If Kn is a Cauchy sequence with respect to . To see the converse let K be given by (6. .44) we infer that Jp (H) is a vector space and the triangle inequality. (6. The two most important cases are p = 1 and p = 2: J2 (H) is the space of Hilbert–Schmidt operators investigated in the previous section and J1 (H) is the space of trace class operators. φj .14.13.126 6. p it is also a Cauchy sequence with respect to . Since C(H) is closed. j Corollary 6. An operator is trace class if and only if it can be written as the product of two Hilbert–Schmidt operators. K = K1 K2 . K is in Jp (H).45) for any ﬁnite ONS. Proof. By CauchySchwarz we have  ϕn .47) and hence K = K1 K2 is trace calls if both K1 and K2 are Hilbert–Schmidt operators.14) and choose K1 = ˆ sj (K) φj . From  ψj . Since the right hand side is independent of the ONS (and in particular on the number of vectors). and we have K in this case. The other requirements for are norm are obvious and it remains to check completeness. Kψn  = n n ∗  K1 ϕn . K1 ϕj p p 1/p + j  ψj . φj respectively K2 = j sj (K) φj . (K1 + K2 )ϕj p j 1/p ≤ j  ψj . Kϕj p j 1/p ≤C (6. K2 ϕj p 1/p ≤ K1 + K2 p (6.
Problem 6.15.6. Now we can also explain the name trace class: Lemma 6. This is even true for all trace class operators and is known as Lidiskij trace theorem (see [17] or [6] for an easy to read introduction). K 1 K 2 ϕn n = n ∗ K 1 ϕn . Clearly for selfadjoint trace class operators. Show that for an orthogonal projection P we have dim Ran(P ) = tr(P ).49) is inﬁnite (for one and hence all ONB by the previous problem). the trace is the sum over all eigenvalues (counted with their multiplicity). K2 Hilbert–Schmidt we have ϕn .m ∗ K2 ψ m . Kϕn (6. Hilbert–Schmidt and trace class operators 127 Proof. show that A p = a p in this case.50) Hence the trace is independent of the ONB and we even have tr(K1 K2 ) = tr(K2 K1 ). . (6. K 1 ψ m = ψ m . To see this you just have to choose the ONB to consist of eigenfunctions. K2 ϕ for every ϕ ∈ H. K1 ϕ ≤ ϕ. Show that A ≥ 0 is trace class if (6. K 2 ϕ n = n. Show that A ∈ Jp ( 2 (N)) if and only if a ∈ p (N).) Problem 6.49) is ﬁnite and independent of the ONB. Let H = 2 (N) and let A be multiplication by a sequence a(n). the trace is linear and if K1 ≤ K2 are both trace class we have tr(K1 ) ≤ tr(K2 ). where we set tr(P ) = ∞ if (6. Moreover. Let {ψn } be another ONB.5. (Hint A is selfadjoint (why?) and A = A A.3. Problem 6.n ∗ K 1 ϕn . then for any ONB {ϕn } the trace tr(K) = n ϕn . K2 Hilbert–Schmidt and use Corollary 6. K 2 K1 ψ m . Proof. If we write K = K1 K2 with K1 .4. Write K = K1 K2 with K1 . K 2 ϕ n ∗ K2 ψ m .11.49) is ﬁnite for one (and √ √ hence all) ONB. Furthermore.3. K 1 ψ m m = = m ϕn . If K is trace class. The rest follows from linearity of the scalar product and K1 ≤ K2 if and only if ϕ. ψ m ψ m . ϕ n m.
f (n) → j∈N k(n + j)f (j) is Hilbert–Schmidt if k(n) ≤ C(n). Hence. we ﬁrst need a good criterion for a point to be in the essential spectrum of A. In this section we want to study the inﬂuence of perturbations on the spectrum. Such a sequence is called singular Weyl sequence. Lemma 6. Moreover. Let ψn be a singular Weyl sequence for the point λ0 . Consider ˜ ˜ ˜ ψn = Pε ψn . Our hope is that at least some parts of the spectrum remain invariant. Next. If λ0 ∈ σd (A) we can ﬁnd an ε > 0 such that Pε = PA ((λ0 − ε. ψn converges weakly to 0.7.12 we have λ0 ∈ σ(A) and hence it suﬃces to show λ0 ∈ σd (A). Let A be selfadjoint. where C(n) is decreasing and summable.16 (Weyl criterion). namely the essential spectrum. if λ0 is in the discrete spectrum. A point λ is in the essential spectrum of a selfadjoint operator A if and only if there is a sequence ψn such that ψn = 1. and (A − λ)ψn → 0. By Lemma 2. where K = √ K ∗ K. . 6. Problem 6.51) Hence our only hope is that the remainder. In fact.4. in general. Proof. Conclude that K p = (tr(Ap ))1/p . Clearly ψn converges weakly to zero and (A − λ0 )ψn → 0. the sequence can chosen to be orthonormal. λ0 + ε)) is ﬁnite rank. Note that if we add a multiple of the identity to A. we can easily remove this eigenvalue with a ﬁnite rank perturbation of arbitrary small norm. we shift the entire spectrum.6. φj . is stable under ﬁnite rank perturbations. we cannot expect a (relatively) bounded perturbation to leave any part of the spectrum invariant. consider A + εPA ({λ0 }). Relatively compact operators and Weyl’s theorem In the previous section we have seen that the sum of a selfadjoint and a symmetric operator is again selfadjoint if the perturbing operator is small. (6.128 6. Show that for K ∈ C we have K = j sj φj . Perturbation theory for selfadjoint operators Problem 6. Show that K : 2 (N) → 2 (N). . To show this.
Then (RA (z)− λ−z )ψn = RA (z) (A−λ)ψn and thus (RA (z)− z−λ 1 → 0. Hence σess (A) ⊆ σess (A + K). The essential spectrum of a selfadjoint operator A is precisely the part which is invariant under rankone perturbations. consider Pn = PA ([λ − n . But ϕn is a sequence of unit length vectors which lives in a ﬁnite dimensional space and converges to 0 weakly. 1 1 n+1 . If RA (z) − RB (z) ∈ C(H) for one z ∈ ρ(A) ∩ ρ(B).6. λ + n ]). where ϕn = RB (z)ψn . Since . if λ0 ∈ σess (A).53) since (A − λ)ψn → 0 by assumption and Kψn → 0 by Lemma 6. suppose λ ∈ σess (A) and let ψn be a corresponding singular 1 Weyl sequence. σess (A) = σ(A + K).λ+ε) 1 ≤ 2 (A − λ0 )ψn 2 (6.4. In particular. ˜ ψn − ψn 2 = ≤ dµψn (λ) R\(λ−ε. Suppose A and B are selfadjoint operators. In fact. (6.17. a contradiction. Lemma 6. Theorem 6. Then ψj ∈ RanPnj . (6. Then ψn converges weakly to zero and hence (A + K − λ)ψn ≤ (A − λ)ψn + Kψn → 0 (6. then σess (A) = σess (B). Moreover. λ − n+1 ) ∪ (λ + rank(Pnj ) > 0 for an inﬁnite subsequence nj . 1 1 Conversely. Reversing the roles of A + K and A shows σess (A + K) = σess (A).55) Proof.K ∗ =K There is even a larger class of operators under which the essential spectrum is invariant.λ+ε) 1 (λ − λ0 )2 dµψn (λ) ε2 R\(λ−ε.8 (iii). Since we have shown that we can remove any point in the discrete spectrum by a selfadjoint ﬁnite rank operator we obtain the following equivalent characterization of the essential spectrum.18 (Weyl).56) (6.54) K∈C(H). Now pick Now let K be a selfadjoint compact operator and ψn a singular Weyl sequence for A. Thus ϕn = ψn −1 ψn is also a singular Weyl sequence.52) ε ˜ ˜ ˜ and hence ψn → 1. Relatively compact operators and Weyl’s theorem 129 Moreover. by our assumption we also have (RB (z) − λ−z )ψn 1 λ−z )ψn → 0 and thus (B − λ)ϕn → 0.
18 applies if B = A + K. Suppose RA (z) − RB (z) ∈ C(H) (6. Proof. where K is relatively compact. For later use observe that set of all operators which are relatively compact with respect to A forms a linear space (since compact operators do) and relatively compact operators have Abound zero. The set of all functions f for which the claim holds is a closed ∗subalgebra of C∞ (R) (with sup norm).59) Let A and B be selfadjoint. Proof. (6. Then the Abound of K is zero. Remember that we have called K relatively compact with respect to A if KRA (z) is compact (for one and hence for all z) and note that the the resolvent diﬀerence RA+K (z) − RA (z) is compact if K is relatively compact.57) for one z ∈ ρ(A) ∩ ρ(B). then this holds for all z ∈ ρ(A) ∩ ρ(B). As a ﬁrst consequence note the following result Theorem 6.19. Hence the claim follows from Lemma 4. Lemma 6. Perturbation theory for selfadjoint operators limn→∞ ϕn = limn→∞ RA (z)ψn = λ − z−1 = 0 (since (RA (z) − 1 1 λ−z )ψn = λ−z RA (z)(A − λ)ψn → 0) we obtain a singular Weyl sequence for B.20.27 the resolvent diﬀerence of two selfadjoint extensions is a ﬁnite rank operator if the defect indices are ﬁnite. the following result is of interest. then all selfadjoint extensions have the same essential spectrum. Suppose A is symmetric with equal ﬁnite defect indices. If the condition holds for one z it holds for all since we have (using both resolvent formulas) RA (z ) − RB (z ) = (1 − (z − z )RB (z ))(RA (z) − RB (z))(1 − (z − z )RA (z )). Let A be selfadjoint and suppose K is relatively compact with respect to A. By Lemma 2. In particular. (6. Theorem 6.21. if A and B are selfadjoint. In addition. Now interchange the roles of A and B.130 6. Lemma 6.4. In addition.58) . then f (A) − f (B) ∈ C(H) for any f ∈ C∞ (R). showing λ ∈ σess (B).
If An converges to A in strong resolvent sense. Show that if −u (x)+q(x)u(x) = zu(x) has a solution for which q∈L u and u are bounded near +∞ (or −∞) but u is not square integrable near +∞ (or −∞). And hence BRA+B (z) = BRA (z)(I + BRA (z))−1 shows that B is also A + B bounded and the result follows from KRA+B (z) = KRA (z)(I − BRA+B (z)) since KRA (z) is compact and BRA+B (z) is bounded.62) 6. (Hint: Use u to construct a Weyl sequence by restricting it to a compact set.60) and observe that the ﬁrst operator is compact and the second is normal and converges strongly to 0 (by the spectral theorem). We say that An converges to A in norm respectively strong resolvent sense if n→∞ lim RAn (z) = RA (z) respectively slim RAn (z) = RA (z) n→∞ (6. then f (An ) converges to f (A) strongly for any bounded continuous function f : Σ → C. Σ = σ(A) ∪ n σ(An ).5.8 (since RA is normal). d Problem 6. Suppose An converges to A in norm resolvent sense. Show that A = − dx2 + q(x).5.) 2 (6. then z ∈ σess (A). Since B is A bounded with Abound less than one. Lemma 6. Proof.8. Strong and norm resolvent convergence 131 Proof. then f (An ) converges to f (A) in norm for any bounded continuous function f : Σ → C with limλ→−∞ f (λ) = limλ→∞ f (λ). Strong and norm resolvent convergence Suppose An and A are selfadjoint operators.3 and the discussion after Lemma 6.61) (6. .6.63) for one z ∈ Γ = C\Σ.22. note the following result which is a straightforward consequence of the second resolvent identity. Suppose A is selfadjoint and B is symmetric with Abound less then one. we can choose a z ∈ C such that BRA (z) < 1. Hence the claim follows from Lemma 6. Now modify your construction to get a singular Weyl sequence by observing that functions with disjoint support are orthogonal. If K is relatively compact with respect to A then it is also relatively compact with respect to A + B. Using the Stone–Weierstraß theorem we obtain as a ﬁrst consequence Theorem 6. Write KRA (λi) = (KRA (i))((A + i)RA (λi)) (6.23. D(A) = H 2 (R) is selfadjoint if ∞ (R). In addition.
and χm (.69) (6. (6. A be selfadjoint operators with D(An ) = D(A). t ≥ 0.68) ≤ (an + bn ) ψ .67) Proof. Let An .25. (6. Then f (An )χm (An ) → f (A)χm (A) by the ﬁrst part and hence (f (An ) − f (A))ψ ≤ f (An ) (1 − χm (An ))ψ (χm (An ) − χm (A))ψ (1 − χm (A))ψ s ∞ + f (An ) + f (A) + (f (An )χm (An ) − f (A)χm (A))ψ (6.132 6. taking adjoints is continuous even with respect to strong convergence) and since it contains f (λ) = 1 and f (λ) = 1 ε λ−z0 this ∗algebra is dense by the Stone–Weierstaß theorem.) → I by Theo As a consequence note that the point z ∈ Γ is of no importance Corollary 6. The usual 3 shows that this ∗algebra is also closed.24. then this holds for all z ∈ Γ. The set of functions for which the claim holds clearly forms a ∗algebra (since resolvents are normal. Lemma 6. Suppose An converges to A in strong resolvent sense. then eitAn → eitA .1.66) Finally we need some good criteria to check for norm respectively strong resolvent convergence.65) (6. m]. and that we have Corollary 6.26. and if all operators are semibounded e−tAn → e−tA . Suppose An converges to A in norm or strong resolvent sense for one z0 ∈ Γ. ψ ∈ D(A) = D(An ). s s t ∈ R. From the second resolvent identity RAn (z) − RA (z) = RAn (z)(A − An )RA (z) we infer (RAn (i) − RA (i))ψ ≤ RAn (i) an RA (i)ψ + bn ARA (i)ψ (6. Then An converges to A in norm resolvent sense if there are sequences an and bn converging to zero such that (An − A)ψ ≤ an ψ + bn Aψ . To see the last claim let χn be a compactly supported continuous funcs tion (0 ≤ χm ≤ 1) which is one on the interval [−m.64) can be made arbitrarily small since f (. Perturbation theory for selfadjoint operators Proof.) ≤ f rem 3.
Similarly. If you wonder why we did not deﬁne weak resolvent convergence. a contradiction. The rest follows from Lemma 1. Using the second resolvent identity we have (RAn (i) − RA (i))ψ ≤ (A − An )RA (i)ψ → 0 (6.27. If An converges to A in strong resolvent sense we have σ(A) ⊆ limn→∞ σ(An ). Let An and A be selfadjoint operators. In particular. then An converges to A in norm resolvent sense. Let An . Proof. λ + ε) = ∅. On the other hand. λ + ε).30. Suppose wlimn→∞ RAn (z) = RA (z) for some z ∈ Γ. λ + 2 ) and vanishes outside (λ − ε. Now what can we say about the spectrum? Theorem 6. Lemma 6. if no domain problems get in the way. strong convergence implies strong resolvent convergence: Lemma 6. Proof. RAn (z ∗ )RAn (z)ψ − RA (z ∗ )RA (z)ψ 1 = ψ.5.11 (iv). Then An converges to A in strong resolvent sense if there there is a core D0 of A such that for any ψ ∈ D0 we have ψ ∈ D(An ) for n suﬃciently large and An ψ → Aψ. norm convergence implies norm resolvent convergence: Corollary 6. Then we can ﬁnd a λ ∈ σ(A) and some ε > 0 such that σ(An ) ∩ (λ − ε. Suppose the ﬁrst claim were wrong. By RAn (z) RA (z) we have also RAn (z)∗ the ﬁrst resolvent identity RA (z)∗ and thus by RAn (z)ψ 2 − RA (z)ψ 2 = ψ. λ + 2 )) implying f (A)ψ = ψ. Proof.29. If An converges to A in norm resolvent sense we have σ(A) = limn→∞ σ(An ). A be selfadjoint operators. Then f (An ) = 0 and hence f (A)ψ = lim f (An )ψ = 0 for every ψ. then also slimn→∞ RAn (z) = RA (z). Strong and norm resolvent convergence 133 and hence RAn (i) − RA (i) ≤ an + bn → 0.13.28. A be bounded selfadjoint operators with An → A. (RAn (z) − RAn (z ∗ ) + RA (z) − RA (z ∗ ))ψ → 0.6.71) z − z∗ Together with RAn (z)ψ RA (z)ψ we have RAn (z)ψ → RA (z)ψ by virtue of Lemma 1. Choose a bounded ε ε continuous function f which is one on (λ − 2 . . since D0 is a core. (6.70) for ψ ∈ (A − i)D0 which is dense. since λ ∈ σ(A) there is a nonzero ψ ∈ RanPA ((λ − ε ε 2 . Let An . here is the answer: it is equivalent to strong resolvent convergence.
Then An converges to 0 in strong resolvent sense. λ1 )) − P ((−∞. λ1 )) = slim PAn ([λ0 . λ]). λ]) = PA ((−∞. Furthermore.λ) by a continuous function f with 0 ≤ f ≤ χ(−∞. λ)))ψ . ((−∞. To overcome this problem let us choose another continuous function g with χ(−∞. λ)) = PA ((−∞.λ] ≤ g ≤ 1. (6. λ1 )) = P ((−∞. λ)))ψ ≤ (PA ((−∞. λ0 ]) we also obtain Corollary 6. Suppose An converges in strong resolvent sense to A. Note that the spectrum can contract if we only have strong resolvent 1 sense: Let An be multiplication by n x in L2 (R). ((−∞.76) .75) (6.72) Proof. which implies RAn (λ + i) < 1 for n suﬃciently large.1) and the same is true for the second if we choose n large (Theorem 6.23). n→∞ n→∞ (6.74) and now all terms are under control. recall that the norm of RA (z) is just one over the distance from the spectrum. λ1 )) = PA ([λ0 . Perturbation theory for selfadjoint operators To see the second claim. The idea is to approximate χ(−∞.73) The ﬁrst term can be made arbitrarily small if we let f converge pointwise to χ(−∞.λ) (Theorem 3. Since we can replace P.λ] ≤ g. If PA ({λ0 }) = PA ({λ1 }) = 0.λ) ≤ χ(−∞. but σ(An ) = R and σ(0) = {0}. which implies λ ∈ σ(An ) for n suﬃciently large.134 6. Then (PA ((−∞.32. the third term can only be made small for ﬁxed n. then slim PAn ((−∞. Using P ((λ0 . In particular.31. n→∞ n→∞ (6. However. λ)))ψ ≤ (g(An ) − f (An ))ψ since f ≤ χ(−∞. So λ ∈ σ(A) implies RA (λ + i) < 1. λ)) = slim PAn ((−∞. Suppose An converges in strong resolvent sense to A. Lemma 6. then slim PAn ((λ0 . λ1 ]). If PA ({λ}) = 0. λ)) by P. λ)) − f (A))ψ + (f (A) − f (An ))ψ + (f (An ) − PAn ((−∞. λ1 ]) = PA ((λ0 . λ ∈ σ(A) if and only if RA (λ + i) < 1.λ) . λ]) in all calculations we are done. (g(An ) − f (An ))ψ ≤ (g(An ) − f (A))ψ + (f (A) − g(A))ψ + (g(A) − g(An ))ψ (6. λ)) − PAn ((−∞. Then (f (An ) − PAn ((−∞.
80) for every bounded continuous f . In fact. The following example shows that the requirement PA ({λ}) = 0 is crucial.6. (6. (6. (6. Problem 6. Suppose An → A in strong resolvent sense and let µψ .9. Strong and norm resolvent convergence 135 Example.78) where {ϕn }n∈N is some (ﬁxed) ONB. Give a counterexample when f is not continuous. Show that f (λ)dµψ.79) d(A. 0]) = I. . 0)) = 0 and P0 ((−∞. Show that for selfadjoint operators. µψ. B) = 2n n∈N 0 0 0 1 . even if we have bounded operators and norm convergence. 0]) = but P0 ((−∞.5. Problem 6. 0)) = PAn ((−∞. let H = C2 and 1 1 0 An = . strong resolvent convergence is equivalent to convergence with respect to the metric 1 (RA (i) − RB (i))ϕn .n be the corresponding spectral measures.n (λ) → f (λ)dµψ (λ) (6.77) n 0 −1 Then An → 0 and PAn ((−∞.10 (Weak convergence of spectral measures).
.
Part 2 Schr¨dinger Operators o .
.
Recall 0 the Schwarz space S(Rn ) = {f ∈ C ∞ (Rn ) sup xα (∂β f )(x) < ∞. For f ∈ C ∞ (Rn ) and α ∈ Nn we set 0 ∂α f = ∂xα1 1 ∂ α f . Let C ∞ (Rn ) be the set of all complexvalued functions which have partial derivatives of arbitrary order. .2) ∞ which is dense in L2 (Rn ) (since Cc (Rn ) ⊂ S(Rn ) is). β ∈ Nn } 0 x (7. . (7.Chapter 7 The free Schr¨dinger o operator 7. ∂n ) is the gradient. ˆ (xα f (x))∧ (p) = iα ∂α f (p). For any multiindex α ∈ Nn and any f ∈ S(Rn ) we have 0 ˆ (∂α f )∧ (p) = (ip)α f (p). 139 .4) Hence we will sometimes write pf (x) for −i∂f (x). where ∂ = (∂1 .1. (7.1. .1) An element α ∈ Nn is called multiindex and α is called its order. n 1 α = α1 + · · · + αn . For f ∈ S(Rn ) we deﬁne 1 ˆ F(f )(p) ≡ f (p) = e−ipx f (x)dn x. · · · ∂xαn n xα = xα1 · · · xαn . .3) (2π)n/2 Rn Then it is an exercise in partial integration to prove Lemma 7. α. (7. The Fourier transform We ﬁrst review some basic facts concerning the Fourier transform which will be needed in the following section.
5) of two functions f. The free Schr¨dinger operator o In particular F maps S(Rn ) into itself. For this the following lemma will be needed. where the branch cut of the root is chosen along the negative real axis. g ∈ S(Rn ) is given by ˆ g (f ∗ g)∧ (p) = (2π)n/2 f (p)ˆ(p). this holds for all z with Re(z) > 0 if we choose the branch cut of the root along the negative real axis. The Fourier transform of the convolution (f ∗ g)(x) = Rn f (y)g(x − y)dn y = Rn f (x − y)g(y)dn y (7. We compute (f ∗ g)∧ (p) = = Rn (7. Proof. ˆ Let φz (x) = exp(−zx2 /2). one can treat each coordinate separately. Next. Proof. z n/2 (7.9) at least for z > 0.140 7.7) where we have used Fubini’s theorem.3. Another useful property is the convolution formula. reducing the problem to the case n = 1. Now we can show .6) 1 (2π)n/2 e−ipx Rn Rn f (y)g(x − y)dn y dn x e−ip(x−y) g(x − y)dn x dn y Rn e−ipy f (y) 1 (2π)n/2 = Rn ˆ g e−ipy f (y)ˆ(p)dn y = (2π)n/2 f (p)ˆ(p). g (7. Lemma 7.1) ˆ zφ z 1 ˆ c = φz (0) = √ 2π 1 exp(−zx2 /2)dx = √ z R (7. Due to the product structure of the exponential. We have e−zx 2 /2 ∈ S(Rn ) for Re(z) > 0 and )(p) = 1 −p2 /(2z) e . Then φz (x)+zxφz (x) = 0 and hence i(pφz (p)+ ˆ (p)) = 0. Thus φz (p) = cφ1/z (p) and (Problem 7. since the integral is holomorphic for Re(z) > 0. Lemma 7.2. However. we want to compute the inverse of the Fourier transform.8) F(e−zx 2 /2 √ Here z n/2 has to be understood as ( z)n .
16) . The Fourier transform 141 Theorem 7. −1.1. Moreover. which implies g(x)(∂α f )(x)dn x = (−1)α Rn Rn f ∈ H r (Rn ). (7. (7. then (F 2 + z 2 )(F + z)(F − z)ψn = (F 4 − z 4 )ψn = (1 − z 4 )ψn → 0 implies z 4 = 1. using Fubini this even implies F 2 (fε )(x) = fε (−x) for any ε > 0. if ψn is a Weyl sequence. Its spectrum satisﬁes σ(F) = {z ∈ Cz 4 = 1} = {1. It remains to compute the spectrum.13) Proof. In fact.7. i. Hence σ(F) ⊆ {z ∈ Cz 4 = 1}.4. The Fourier transform F : S(Rn ) → S(Rn ) is a bijection. (7. −i}.12) for f ∈ S(Rn ) and thus we can extend F to a unitary operator: Theorem 7. Its inverse is given by 1 F −1 (g)(x) ≡ g (x) = ˇ eipx g(p)dn p.14) We will abbreviate ˆ ∂α f = ((ip)α f (p))∨ . The Fourier transform F extends to a unitary operator F : L2 (Rn ) → L2 (Rn ).11) Since limε↓0 fε (x) = f (x) for every x ∈ Rn (Problem 7.5. where we will explicitly compute an orthonormal basis of eigenfunctions. It suﬃces to show F 2 (f )(x) = f (−x).10) (2π)n/2 Rn We have F 2 (f )(x) = f (−x) and thus F 4 = I. We defer the proof for equality to Section 8. From Fubini’s theorem we also obtain Parseval’s identity 1 ˆ ˆ f (p)2 dn p = f (x)∗ f (p)eipx dn p dn x n (2π)n/2 Rn Rn R = Rn f (x)2 dn x (7. (7. Let us introduce the Sobolev space ˆ H r (Rn ) = {f ∈ L2 (Rn )pr f (p) ∈ L2 (Rn )}. where fε (x) = 1 εn Rn φ1/ε2 (x − y)f (y)dn y.2).3. we infer from dominated convergence F 2 (f )(x) = limε↓0 F 2 (fε )(x) = limε↓0 fε (−x) = f (−x). Consider φz (x) from the proof of the previous lemma and observe F 2 (φz )(x) = φz (−x). (7.1 also allows us to extend diﬀerentiation to a larger class. Proof. α ≤ r (7.15) (∂α g)(x)f (x)dn x. Lemma 7.
18) ˆ shows f ∈ C∞ (Rn ) for arbitrary f ∈ L1 (Rn ) since S(Rn ) is dense in L1 (Rn ). The corresponding non relativistic Hamilton operator. n = N d. Extend Lemma 0. (7. Let C∞ (Rn ) denote the Banach space of all continuous functions f : Rn → C which vanish at ∞ equipped with the sup norm. The free Schr¨dinger operator o for f ∈ H r (Rn ) and g ∈ S(Rn ).20) Our ﬁrst task is to ﬁnd a good domain such that H0 is a selfadjoint operator. Lemma 7.6 (RiemannLebesgue).2. More generally. ∂α f is the derivative of f in the sense of distributions.19) ∆= j=1 ∂2 . D(H0 ) = H 2 (Rn ). ∂x2 j (7. The free Schr¨dinger operator o In Section 2. That is.32 to the case u.142 7.17). (7. if the particles do not interact. Clearly we have f ∈ C∞ (Rn ) if f ∈ S(Rn ). where ∆ is the Laplace operator n (7. Show that R exp(−x2 /2)dx = gral and evaluate it using polar coordinates. By Lemma 7.21) .1 we have that ˆ − ∆ψ(x) = (p2 ψ(p))∨ (x). (Hint: Square the inte Problem 7. the Hilbert space for N particles in Rd is L2 (Rn ). Problem 7.2.22) ψ ∈ H 2 (Rn ).1 we have seen that the Hilbert space corresponding to one particle in R3 is L2 (R3 ). the estimate ˆ sup f (p) ≤ p 1 sup (2π)n/2 p e−ipx f (x)dn x = Rn 1 (2π)n/2 f (x)dn x Rn (7. 7. and hence the operator H0 ψ = −∆ψ. f ∈ S(Rn ) (compare Problem 0. Then the Fourier transform is a bounded map from L1 (Rn ) into C∞ (Rn ) satisfying ˆ f ∞ ≤ (2π)−n/2 f 1 . Moreover. (7. is given by H0 = −∆. we have the RiemannLebesgue lemma.1. Finally.17) ˆ Proof.) √ 2π.
Rp2 (z)ψ = where d˜ψ (r) = χ[0. S n−1 Rn ˆ ψ(p)2 n d p= p2 − z R r2 1 d˜ψ (r). RH0 (z)ψ = R λ−z where 1 dµψ (λ) = χ[0. ψ. √ ˆ ψ( λω)2 dn−1 ω dλ. S n−1 (7. after a change of coordinates. The free Schr¨dinger operator o 143 is unitarily equivalent to the maximally deﬁned multiplication operator (F H0 F −1 )ϕ(p) = p2 ϕ(p). we have 1 dµψ (λ). we note that the compactly supported smooth functions are a core for H0 .27) (7.3) and hence it ∞ suﬃces to show that the closure of H0 Cc (Rn ) contains H0 S(Rn ) . σsc (H0 ) = σpp (H0 ) = ∅. (7. (7. It is not hard to see that S(Rn ) is a core (Problem 7. The set Cc (Rn ) = {f ∈ S(Rn )supp(f ) is compact} is a core for H0 .8. Proof.25) µ −z (7. Show that S(Rn ) is a core for H0 . RH0 (z)ψ = ψ. ∞). First observe that ˆ ˆ ψ.3. (7. ∞ Lemma 7. It suﬃces to show that dµψ is purely absolutely continuous for every ψ.2.∞) (λ)λn/2−1 2 proving the claim.∞) (r)rn−1 µ ˆ ψ(rω)2 dn−1 ω dr. Note also that the quadratic form of H0 is given by n qH0 (ψ) = j=1 ∂j ψ(x)2 dn x. The free Schr¨dinger operator H0 is selfadjoint and its o spectrum is characterized by σ(H0 ) = σac (H0 ) = [0.7. Rn ψ ∈ Q(H0 ) = H 1 (Rn ).28) Finally.7.29) Problem 7.24) Proof. (Hint: Show that the closure of H0 S(Rn ) contains H0 . Set let ϕ(x) ∈ Cc 1 ∞ ϕn (x) = ϕ( n x). D(p2 ) = {ϕ ∈ L2 (Rn )p2 ϕ(p) ∈ L2 (Rn )}. then ψn (x) = ϕn (x)ψ(x) is in Cc (Rn ) for every ψ ∈ S(Rn ) and ψn → ψ respectively ∆ψn → ∆ψ.23) Theorem 7.) .26) Hence. (7. To see this. ∞ (Rn ) which is one for x ≤ 1 and vanishes for x ≥ 2.
since 2 e−itp is not in L2 .31) Then fε (H0 )ψ → e−itH0 ψ by Theorem 3. Thus we have spreading of wave functions in this case. (7.34) by the RiemannLebesgue lemma.33) for t = 0 and ψ ∈ L1 ∩ L2 . Observe e−itH0 ψ(x) = = 2 ei 4t (4πit)n/2 1 2it n/2 x2 ei 4t ψ(y)ei 2t dn y Rn y2 xy ei 4t x2 ei 4t ψ(y) y2 ∧ ( x ).3. it is not hard to draw some immediate consequences. if ψ ∈ L2 (Rn ) ∩ L1 (Rn ). However. Consider fε (p2 ) = e−(it+ε)p .35) Moreover. For example.32) ψ(y)dn y (7. 2t (7. 2 ε > 0. 7. since exp(i y )ψ(y) → ψ(y) in L2 as t → ∞ (dominated conver4t gence) we obtain . We have 2 ˆ e−itH0 ψ(x) = F −1 e−itp ψ(p). The free Schr¨dinger operator o Problem 7.144 7. it is even possible to determine the asymptotic form of the wave function for large t as follows.4. For general ψ ∈ L2 the integral has to be understood as a limit. Moreover.3 and the convolution formula we have fε (H0 )ψ(x) = and hence e−itH0 ψ(x) = 1 (4πit)n/2 ei Rn x−y2 4t 1 (4π(it + ε))n/2 e Rn − 4(it+ε) x−y2 ψ(y)dn y (7. a more careful analysis is needed. Moreover. by Lemma 7. The time evolution in the free case Now let us look at the time evolution. then ψ(t) ∈ C(Rn ) for t = 0 (use dominated convergence and continuity of the exponential) and satisﬁes ψ(t) ∞ ≤ 1 ψ(0) 4πtn/2 1 (7. Using this explicit form.30) The right hand side is a product and hence our operator should be expressible as an integral operator via the convolution formula.1. Show that {ψ ∈ S(R)ψ(0) = 0} is dense but not a core for d2 H0 = − dx2 . (7.
We will try to use a similar approach as for the time evolution in the previous section. Proof.7. By symmetry it suﬃces to consider g(x)f (p).4.9. The resolvent and Green’s function Now let us compute the resolvent of H0 . If f. a particle will escape to inﬁnity. Next we want to apply the RAGE theorem in order to show that for any initial condition.37) 1 2it n/2 x ˆ x ei 4t ψ( ) → 0 2t 2 (7. Let g(x) be the multiplication operator by g and let f (p) be ˆ the operator given by f (p)ψ(x) = F −1 (f (p)ψ(p))(x). For any ψ ∈ L2 (Rn ) we have e−itH0 ψ(x) − in L2 as t → ∞.4. . g vanish at inﬁnity. In particular. The resolvent and Green’s function 145 Lemma 7. we will use a small ruse.) 7.39) lim χΩ e−itH0 ψ 2 =0 (7. Thus gR (x)fR (p) is compact and tends to g(x)f (p) in norm since f. However. this lemma implies that χΩ (H0 + i)−1 is compact if Ω ⊆ Rn is bounded and hence t→∞ (7. Lemma 7.36) are compact if f. g ∈ L2 . g ∈ L2 (Rn ). g are bounded then the functions fR (p) = χ{pp2 ≤R} (p)f (p) and gR (x) = χ{xx2 ≤R} (x)g(x) are in L2 .38) ˇ shows that g(x)f (p) is Hilbert–Schmidt since g(x)f (x − y) ∈ L2 (Rn × Rn ). Denote by L∞ (Rn ) the ∞ bounded Borel functions which vanish at inﬁnity. then g(x)f (p)ψ(x) = 1 (2π)n/2 ˇ g(x)f (x − y)ψ(y)dn y Rn (7.34). Let f. since it is highly nontrivial to compute the inverse Fourier transform of exp(−εp2 )(p2 − z)−1 directly. In other words. Then f (p)g(x) and g(x)f (p) (7.40) for any ψ ∈ L2 (Rn ) and any bounded subset Ω of Rn . g ∈ L∞ (Rn ) and (extend to) Hilbert–Schmidt operators if ∞ f. the particle will eventually escape to inﬁnity since the probability of ﬁnding the particle in any bounded set tends to zero.10. (If ψ ∈ L1 (Rn ) this of course also follows from (7.
46) ν=0 ν=0 (7. (7. (4πt)n/2 r > 0.43) G0 (z. G0 (z. r) is called Green’s function of H0 . In particular. e−tH0 ψ(x) = 1 (4πt)n/2 e− Rn x−y2 4t ψ(y)dn y. Re(z) < 0. The free Schr¨dinger operator o Note that ∞ RH0 (z) = 0 ezt e−tH0 dt. (7. RH0 (z)ψ for Re(z) < 0 it is equal to this function for all z ∈ ρ(H0 ). by Fubini. we have RH0 (z)ψ(x) = where ∞ G0 (z. The integral can be evaluated in terms of modiﬁed Bessel functions of the second kind 1 G0 (z.45) The functions Kν (x) satisfy the following diﬀerential equation d2 1 d ν2 + −1− 2 dx2 x dx x and have the following asymptotics Kν (x) = for x → 0 and Kν (x) = π −x e (1 + O(x−1 )) 2x (7. (7. Hence. Moreover. ψ ∈ S(Rn ) (by Morea’s theorem). 2 (7. x − y)ψ(y)dn y. . Re(z) < 0 (7. r) = 0 r2 1 e− 4t +zt dt. t > 0. x − y)ψ(y)dn ydn x Rn Rn (7. since both functions are analytic in this domain. Rn (7.49) is analytic for z ∈ ρ(H0 ) and ϕ.42) by the same analysis as in the previous section. Since it is equal to ϕ. r) has an analytic continuation for z ∈ C\[0. For more information see for example [22].43) for all z ∈ ρ(H0 ) such that ϕ(x)G0 (z.48) x −ν + O(x−ν+1 ) 2 x − ln( 2 ) + O(1) Γ(ν) 2 Kν (x) = 0 (7.47) for x → ∞.41) by Lemma 4. Hence we can deﬁne the right hand side of (7.43) holds for all z ∈ ρ(H0 ).44) The function G0 (z. r) = 2π −z 4π 2 r2 n−2 4 √ K n −1 ( −zr).1.146 7. In particular. ∞) = ρ(H0 ).
(7. n = 3. we have the case of spherical Bessel functions which can be expressed in terms of elementary functions.43) directly in the case n = 1. For example. .51) G0 (z. n = 1.50) 2 −z and 1 −√−z r . (7.4. we have √ 1 G0 (z.7.5. Verify (7. The resolvent and Green’s function 147 If n is odd. r) = √ e− −z r . r) = e 4πr Problem 7.
.
Moreover. Then ϕi (x1 )ϕj (x2 )ϕk (x3 ) is an orthonormal basis for L2 (R3 ) and x1 can be written as a orthogonal sum of operators restricted to the subspaces spanned by ϕj (x2 )ϕk (x3 ). Hence the claim follows (or use Theorem 4.2) 1 ≤ j ≤ 3. there are several other important observables associated with a single particle in three dimensions. For ψ ∈ S(R3 ) we compute lim i e−itxj ψ(x) − ψ(x) = xj ψ(x) t (8.12). 1 ≤ j ≤ 3.1.3) 149 . many important consequences about these observables can be derived. it is not hard to see that the spectrum of xj is purely absolutely continuous and given by σ(xj ) = R. It is custom to combine all three operators to one vector valued operator x. let ϕ(x) be an orthonormal basis for L2 (R). Position and momentum Apart from the Hamiltonian H0 .3 it is essentially selfadjoint on ψ ∈ S(R3 ). which corresponds to the kinetic energy. Each subspace is unitarily equivalent to L2 (R) and x1 is given by multiplication with the identity. By Corollary 5. Next. (8.Chapter 8 Algebraic methods 8. First consider the oneparameter unitary group (Uj (t)ψ)(x) = e−itxj ψ(x). consider the oneparameter unitary group of translations (Uj (t)ψ)(x) = ψ(x − tej ). Using commutation relation between these observables. which is known as position operator. In fact. (8.1) t→0 and hence the generator is the multiplication operator by the jth coordinate function.
since it is unitarily equivalent to xj by virtue of the Fourier transform we conclude that the spectrum of pj is again purely absolutely continuous and given by σ(pj ) = R. (8. B]) (8. The Weyl relations also imply that the meansquare deviation of position and momentum cannot be made arbitrarily small simultaneously: Theorem 8. pj ψ(t) = 0. B]. ψ ∈ S(R3 ). (8. ψ(t) = e−itH0 ψ(0) ∈ S(R3 ). B} + [A. B} = AB + B A 2 2 ˆ ˆ where {A. then for any ψ ∈ D(AB) ∩ D(AB) we have 1 ∆ψ (A)∆ψ (B) ≥ Eψ ([A.5) d ψ(t).7) which is known as the Weyl relation. Let us ﬁx ψ ∈ D(AB) ∩ D(AB) and abbreviate ˆ A = A − Eψ (A). ABψ 2 =  ψ. Algebraic methods where ej is the unit vector in the jth coordinate direction.4) = t→0 t i ∂xj ∂ and hence the generator is pj = 1 ∂xj . [A.12) (8. {A. ∆ψ (B) = Bψ and hence by CauchySchwarz ˆ ˆ  Aψ. Moreover.8) 2 with equality if (B − Eψ (B))ψ = iλ(A − Eψ (A))ψ. Now note that 1 ˆ ˆ 1 ˆˆ ˆ ˆ ˆˆ ˆˆ AB = {A. xk ]ψ(x) = δjk ψ(x). Suppose A and B are two symmetric operators. Bψ  ≤ ∆ψ (A)∆ψ (B). B}ψ 2 +  ψ. So 1 1 ˆ ˆ ˆˆ ˆ ˆ  Aψ.6) dt that is. pj ]ψ(x) = 0.13) . The operator p is known as momentum operator. (8. we have ψ ∈ S(R3 ) (8.11) λ ∈ R\{0}. ˆ B = B − Eψ (B). For ψ ∈ S(R3 ) we compute ψ(x − tej ) − ψ(x) 1 ∂ lim i ψ(x) (8. the momentum is a conserved quantity for the free motion. Similarly one has [pj . Note that since [H0 .150 8. Proof. Bψ 2 =  ψ. (8. B] are symmetric.10) ˆ ˆ Then ∆ψ (A) = Aψ .1 (Heisenberg Uncertainty Principle).9) (8. B]ψ 2 2 2 (8. {A. or if ψ is an eigenstate of A or B. B} and i[A. Again it is essentially selfadjoint i on ψ ∈ S(R3 ).
Angular momentum 151 which proves (8. 1 ≤ j ≤ 3. (8. Lj ψ(t) = 0. {A.k=1 (8.1. Lj ]ψ(x) = 0.8. and 8. ψ ∈ S(R3 ). j. we see that the spectrum is a subset of Z.15) λ 2 which satisfy Eψ (x) = x0 and Eψ (p) = p0 respectively ∆ψ (pj )2 = 1 ∆ψ (xk )2 = 2λ . B}ψ = 0. (8. Since ei2πLj = I.19) we have again d ψ(t).8). Problem 8. Inserting the ﬁrst into the second ∗ ) Aψ 2 and shows Re(z) = 0. Check that (8. the continuous spectrum is empty. ˆ requirement gives 0 = (z − z In case of position and momentum we have ( ψ = 1) δjk ∆ψ (pj )∆ψ (xk ) ≥ 2 and the minimum is attained for the Gaussian wave packets ψ(x) = λ π n/4 (8. In particular. the angular momentum is a conserved quantity for the free motion as well. We will show below that we have σ(Lj ) = Z. (8.17) Again one combines the three components to one vector valued operator L = x ∧ p. (8.15) realizes the minimum.14) e− 2 x−x0  λ 2 −ip x 0 .20) dt that is. Angular momentum Now consider the oneparameter unitary group of rotations (Uj (t)ψ)(x) = ψ(Mj (t)x). Note that since [H0 .18) 3 εijk xj pk ψ(x). = 0 else (8. For ψ ∈ S(R3 ) we compute ψ(Mi (t)x) − ψ(x) lim i = t→0 t where εijk if ijk is an even permutation of 123 1 −1 if ijk is an odd permutation of 123 . .2. which is known as angular momentum operator. ψ(t) = e−itH0 ψ(0) ∈ S(R3 ).16) where Mj (t) is the matrix of rotation around ej by an angle of t. ˆ ˆ To have equality if ψ is not an eigenstate we need Bψ = z Aψ for equality ˆ ˆ in CauchySchwarz and ψ.2.
(8. (All commutation relations below hold for ψ ∈ S(R3 ). L2 and L3 are given by ﬁnite matrices on Dk .21). L2 ψ = l(l + 1)ψ. ψ ∈ S(R3 ). ψ ∈ S(R3 ).) Denote by Hl. Algebraic methods Moreover.2. Now let us try to draw some further consequences by using the commutation relations (8. Dk is invariant under L2 and L3 and hence Dk reduces L2 and L3 . Proof.m the set of all functions in D satisfying L3 ψ = mψ. . Introducing L2 = L2 + L2 + L2 it is straightforward to check 1 2 3 [L2 . xj }. In this respect the following domain x2 D = span{xα e− 2  α ∈ Nn } ⊂ S(Rn ) (8. pj .3.152 8.22) 0 is often used. Increasing k we get an orthonormal set of simultaneous eigenfunctions whose span is equal to D. Lemma 8.9 it suﬃces to consider the case n = 1. Hence there is an orthonormal basis of simultaneous eigenfunctions of L2 and L3 .26) . which Since it is invariant under the unitary groups generated by Lj . It has the nice property that the ﬁnite dimensional subspaces Dk = span{xα e− x2 2  α ≤ k} (8. Then 1 √ 2π ϕ(x)e −x 2 2 k j=1 (itx)j =0 j! x2 2 (8. the space Pk Hm is invariant under L2 which shows that we can choose an orthonormal basis consisting of eigenfunctions of L2 for Pk Hm . Now let Hm = Ker(L3 − m) and denote by Pk the projector onto Dk .25) Moreover. Lj ]ψ(x) = 0. The subspace D ⊂ L2 (Rn ) deﬁned in (8. Hence ϕ(x) = 0. Since L2 and L3 commute on Dk . Kj ∈ {Lj . By Lemma 1. But the limit is the Fourier transform of ϕ(x)e− shows that this function is zero. ψ = 0 for every ψ ∈ D.21) and these algebraic commutation relations are often used to derive information on the point spectra of these operators. Suppose ϕ. In particular. (8. we even have 3 [Li .23) are invariant under Lj (and hence they reduce Lj ). the operators Lj are essentially selfadjoint on D by Corollary 5.22) is dense.24) for any ﬁnite k and hence also in the limit k → ∞ by the dominated convergence theorem. (8. Kj ]ψ(x) = i k=1 εijk Kk ψ(x).
L L± ψ = (l(l + 1) − m(m ± 1)) ψ for every ψ ∈ Hl.m = 1. Moreover.27) (8.3. L2 (L± ψ) = l(l + 1)(L± ψ). There exists an orthonormal basis of simultaneous eigenvectors for the operators L2 and Lj . First introduce two new operators L± = L1 ± iL2 .35) We will rederive this result using diﬀerent methods in Section 10. (l − m)!(2l)! − (8.3. since L2 = L2 ± L3 + L L± 3 we obtain L± ψ 2 [L3 . we note that (8. Up to this point we know σ(L2 ) ⊆ {l(l + 1)l ∈ N0 }. First of all we observe x2 1 ψ0.30) = ψ.0 . [L± . .m = {0} for l ∈ N0 . L± Hl. Moreover.31) π Next. Hence we must have Hl. σ(L3 ) = Z.l . [L . In summary. their spectra are given by σ(L2 ) = {l(l + 1)l ∈ N0 }.29) (8. Then.l (x) = √ (x1 ± ix2 )l ψ0. (8. σ(L3 ) ⊆ Z.0 (x) = 3/2 e− 2 ∈ H0. (8.m we have L3 (L± ψ) = (m ± 1)(L± ψ). x± ] = 2x± (1 ± L3 ) Hence if ψ ∈ Hl. then (x1 ± ix2 )ψ ∈ Hl±1. 2x3 L± . for every ψ ∈ Hl. l.33) (8. In order to show that equality holds in both cases. And thus 1 ψl. .m = {0} for m > l. (8. x± ] = ±x± .m → Hl.m±1 is injective unless m = l. x± ] = 0.m (x) = (l + m)! Ll−m ψl. −l + 1.2.32) [L± .8. l! respectively ψl. x ] = ±2x3 . l − 1. L± Hl.l±1 . we need to show that Hl.28) that is.m = {0} for l ∈ N0 . Moreover.l . Angular momentum 153 By L2 ≥ 0 and σ(L3 ) ⊆ Z we can restrict our attention to the case l ≥ 0 and m ∈ Z. . .l (x) ∈ Hl. . (8.34) The constants are chosen such that ψl.m → Hl. Theorem 8. L± ] = ±L± .m . If ψ = 0 we must have l(l + 1) − m(m ± 1) ≥ 0 which shows Hl.m . m = −l.0 (x) ∈ Hl. 2 x± = x1 ± ix2 .m±1 .21) implies [L3 . (8.
The harmonic oscillator Finally. (8. A+ ψ 2 = ψ. Introducing 1 √ 1 d A± = √ ωx √ ω dx 2 we have [A− . Moreover.42) e 2 ∈ D. A− A+ ψ = (n + 1) ψ 2 respectively A− ψ 2 = n ψ 2 in this case and hence we conclude that σ(N ) ⊆ N0 If N ψ0 = 0. dxn (8.3. Moreover. Algebraic methods 8.44) 2n n! π where Hn (x) is a polynomial of degree n given by Hn (x) = e x2 2 x2 2 ω > 0. A± ] = ±A± . the harmonic oscillator H = H0 + ω 2 x2 .45) we conclude span{ψn } = D. . since [N. The polynomials Hn (x) are called Hermite polynomials.  α ∈ N3 } ⊆ L2 (R3 ) 0 (8. for any function in D.154 8. Moreover. since √ ωx2 1 ω 1/4 ψn (x) = √ Hn ( ωx)e− 2 (8. As domain we will choose D(H) = D = span{xα e− from our previous section. then we must have A− ψ = 0 and the normalized solution of this last equation is given by ω 1/4 − ωx2 ψ0 (x) = (8.40) N = A+ A− .41) we see that N ψ = nψ implies N A± ψ = (n ± 1)A± ψ.36) (8.38) (8. A+ ] = 1 and H = ω(2N + 1). x− d dx n e− x2 2 = (−1)n ex 2 dn −x2 e . We will ﬁrst consider the onedimensional case. In summary.39) (8. π Hence 1 ψn (x) = √ An ψ0 (x) (8. let us consider another important model whose algebraic structure is similar to those of the angular momentum.37) (8.43) + n! is a normalized eigenfunction of N corresponding to the eigenvalue n.
50) Problem 8.44). + + n! n! since Fψ0 = ψ0 by Lemma 7.49) (8. In particular.3. by FA± = iA± F we even infer (−i)n 1 Fψn = √ FAn ψ0 = √ An Fψ0 = (−i)n ψn . σ(F) = {z ∈ Cz 4 = 1}. there is also a close connection with the Fourier transformation. without restriction we choose ω = 1 and consider only one dimension. (8. d2 dx2 (8.8. Show that H = − + q can be written as H = AA∗ .46) Finally. if the diﬀerential equation ψ + qψ = 0 has a positive solution.4.48) (8. where d A = − dx + φ. Moreover.3.47) (8. The harmonic oscillator 155 Theorem 8. ˜ Compute H = A∗ A.2. Then it easy to verify that H commutes with the Fourier transformation FH = HF. with ψnj (xj ) from (8. The spectrum is given by σ(H) = {(2n + 3)ωn ∈ N0 }.n2 . The harmonic oscillator H is essentially self adjoint on D and has an orthonormal basis of eigenfunctions ψn1 .) ψ .n3 (x) = ψn1 (x1 )ψn2 (x2 )ψn3 (x3 ). (Hint: φ = ψ .
.
the case of a particle in three dimensions can in some situations be reduced to the investigation of SturmLiouville equations. we will see how this works when explicitly solving the hydrogen atom. f. realvalued loc (ii) q ∈ L1 (I). realvalued loc (iii) r ∈ L1 (I). We require (i) p−1 ∈ L1 (I).1. τ f (x) = 1 r(x) − d d p(x) f (x) + q(x)f (x) . the SturmLiouville equations.Chapter 9 One dimensional Schr¨dinger operators o 9. dx dx f. pf ∈ ACloc (I) (9.2) where I = (a. The suitable Hilbert space is b L2 ((a.1) The case p = r = 1 can be viewed as the model of a particle in one dimension in the external potential q. SturmLiouville operators In this section we want to illustrate some of the results obtained thus far by investigating a speciﬁc example. In particular. Moreover. (9. r(x)dx). g = a f (x)∗ g(x)r(x)dx. positive loc 157 . b). b) ⊂ R is an arbitrary open interval.
f ) − Wa (g ∗ . Suppose rg ∈ L1 (I).) p ∈ ACloc (I). (9. (9.158 9. then there exists a unique solution loc f. f is holomorphic with respect to z.g. f . q ∈ ∞ r−1 ∈ L∞ (I) since C0 (I) ⊂ D(τ ) in this case. g ∈ D(τ ). r dx) is given by D(τ ) = {f ∈ L2 (I. g ∈ D(τ ) in (9. f ) − Wc (g ∗ .b (g ∗ . which results in g.3) L2 (I).1) is called regular at a. c)) (c ∈ I). Using integration by parts (twice) we obtain (a < c < d < b): d c g ∗ (τ f ) rdy = Wd (g ∗ . Finally.9) z ∈ C. r dx)f. β ∈ C.1.4). u2 ) = W (u1 . then the SturmLiouville equation (9. r ∈ L1 ((a. τ f ∈ L2 (I. loc It is not clear that D(τ ) is dense unless (e. it is nonzero if and only if u1 and u2 are linearly independent (compare Theorem 9.4 below. If we choose f.7) (9. pf . pf ∈ ACloc (I) of the diﬀerential equation (τ − z)f = g. pf ∈ ACloc (I). than we can take the limits c → a and d → b. If it is both regular at a and b it is called regular. c ∈ I. Theorem 9. pg ∈ ACloc (I) where Wx (f1 .1 below).2 . α.5) In addition. τ u1. One dimensional Schr¨dinger operators o If a is ﬁnite and if p−1 . We will defer the general loc case to Lemma 9. u2 ). we perform a little calculation. τ f = Wb (g ∗ . q. satisfying the initial condition f (c) = α. (9. g. we recall the following wellknown result from ordinary diﬀerential equations. .6) Moreover. f ) + c d (τ g)∗ f rdy. Since we are interested in selfadjoint operators H associated with (9.4) for f.2 = zu1.4) also shows that the Wronskian of two solutions of τ u = zu is constant Wx (u1 . Note that f. The maximal domain of deﬁnition for τ in L2 (I. Here Wa. Equation (9. (9. Similarly for b. f ) + τ g. f ) has to be understood as limit. r dx)}.8) f. p . (9. pf can be extended continuously to a regular end point. (pf )(c) = β. f2 ) = p(f1 f2 − f1 f2 ) (x) is called the modiﬁed Wronskian. (9.1).
∀g ∈ D(τ )}.11) = (q − z)u1 α + = (q − z)f − gr which proves the claim.10) gives the second. j=0 Proof. . l1 . Then l = j=1 n αj lj for some constants αj ∈ C. For this the following elementary fact will be needed.13) . This deﬁnition clearly ensures that the Wronskian of two such functions vanishes on the boundary. . u2 )gr u1 g dy − gr (9. . . Our ﬁrst task is to compute the closure of A0 and its adjoint. It suﬃces to check τ f − z f = g. Then the map L : V → Cn .4. The operator A0 is densely deﬁned and its closure is given by A0 f = τ f. Diﬀerentiating the ﬁrst equation of (9. Hence j=1 there are vectors fk ∈ V such that lj (fk ) = 0 for j = k and lj (fj ) = 1.2. u2 ) = 1. Next we compute (pf ) = (pu1 ) α + u2 gr dy + (pu2 ) β − u1 gr dy − W (u1 . ln are linear functionals (deﬁned on all of V ) such that n Ker(lj ) ⊆ Ker(l).1 if u1 (c) = (pu2 )(c) = 1 and (pu1 )(c) = u2 (c) = 0.1. β coincide with those from Theorem 9. (9. u2 are two solutions of (τ −z)u = 0 with W (u1 . g) = Wb (f. . ln (f )) is surjective (since x ∈ Ran(L)⊥ implies n xj lj (f ) = 0 for all f ). implying that A0 is symmetric.3. SturmLiouville operators 159 Lemma 9.10) c f (x) = u1 (x) α + c u2 gr dy + u2 (x) β − Note that the constants α.12) where ACc (I) are the functions in AC(I) with compact support.9. Now we are ready to prove Lemma 9. Proof. Suppose u1 . (9. u2 gr dy + (q − z)u2 β − Now we want to obtain a symmetric operator and hence we choose A0 f = τ f. D(A0 ) = {f ∈ D(τ )  Wa (f. . Then any other solution of (9. Thus j=1 j=1 j=1 we can choose αj = l(fj ). Suppose V is a vector space and l. g) = 0. D(A0 ) = D(τ ) ∩ ACc (I).8) can be written as (α. First of all it is no restriction to assume that the functionals lj are linearly independent. Then f − n lj (f )fj ∈ n Ker(lj ) and hence l(f ) − n lj (f )l(fj ) = 0. . (9. f → (l1 (f ). β ∈ C) x x f (x) = u1 (x) α + c x u2 gr dy + u2 (x) β − c x u1 gr dy . Lemma 9. . u1 gr dy .
In particular. So suppose we have two choices k1 = k2 . u2 ) = 0.15) ˜ ˜ for some k ∈ L2 (I.3 implies b a ˜ (h(x) − h(x) + α1 u1 (x) + α2 u2 (x))∗ g(x)r(x)dx = 0.19) ˜ and hence h = h + α1 u1 + α2 u2 ∈ D(τ ).2 u2 = ˜ ˜ ˜ h2 + α2. h1 − h2 is in the kernel of τ and hence ˜ 1 = τ h2 = k2 contradiction our assumption. Now Lemma 9. rdx) c (9.3.17) on L2 (I. Denote the set on the right hand side of (9.13) by D. ˜ k1 = τ h Next we turn to A0 . A0 f = k. ∀f ∈ D(A0 ). where uj are two solutions of τ u = 0 with W (u1 .7). By (9. ∀f ∈ D(A0 ) (9. h) = 0. Conversely.7) to conclude Wa (f. In fact. One dimensional Schr¨dinger operators o Its adjoint is given by A∗ f = τ f. there are ˜ ˜ ˜ corresponding functions h1 and h2 such that h = h1 + α1. c We have Ker(l1 ) ∩ Ker(l2 ) ⊆ Ker(l).14) Proof.16). f . r dx).20) . 0 (9. h) + Wb (f. since A0 ⊆ A∗ 0 0 we can use (9. 0 (9. f ∈ D(A0 ). Using (9.7) we have D(τ ) ⊆ D(A∗ ) and it remains to show D(A∗ ) ⊆ D(τ ).18) is in D(A0 ) and g = τ f ∈ Ker(l) by (9. Then we have D ⊆ D(A∗∗ ) = A0 by (9. b lj (g) = a uj (x)∗ g(x)r(x)dx. then x b f (x) = u1 (x) a u2 (y)g(y)r(y)dy + u2 (x) x u1 (y)g(y)r(y)dy (9.10) we can ﬁnd a h such that τ h = k and from integration by parts we obtain b a ˜ (h(x) − h(x))∗ (τ f )(x)r(x)dx = 0. r dx). h ∈ D(A∗ ).2 u2 .1 u1 + α2. (9.16) ˜ Clearly we expect that h − h will be a solution of the τ u = 0 and to prove this we will invoke Lemma 9. 0 0 If h ∈ D(A∗ ) we must have 0 h. ∀g ∈ L2 (I. Therefore we consider the linear functionals b l(g) = a ˜ (h(x) − h(x))∗ g(x)r(x)dx. (9. We start by computing A∗ and ignore the fact that we don’t know 0 wether D(A0 ) is dense for now. 0 D(A∗ ) = D(τ ).160 9. Then by the above calculation.1 u1 + α1. Now what if D(A0 ) were not dense? Then there would be some freedom in choice of k since we could always add a component in D(A0 )⊥ . if g ∈ Ker(l1 ) ∩ Ker(l2 ).
limit point alternative 161 ˜ Now replace h by a h ∈ D(A∗ ) which coincides with h near a and vanishes 0 ˜ ˜ identically near b (Problem 9. (pg )(a) for g ∈ D(τ ) arbitrarily.p. Choosing f1 = f. f3 ) = 0 (9. Wb (f. D(B) = {f ∈ L2 (0. This follows since we can prescribe the values of g(a). then we have ˆa with W (v Wa (v. f ) = 0 and Wa (v. (Hint: Lemma 9. f4 ) + Wx (f1 . Finally. Example. f ) = Wa (v. f4 ∈ D(τ ) we have the Pl¨cker identity u Wx (f1 .18) 2 9. D(A0 ) = {f ∈ H 2 [0.2. . (9.22).2) d Problem 9. it is limit circle at a. Hence we collect a few properties ﬁrst.21) Proof. Then Wa (f. f3 = v. f4 = f we infer (9. Problem 9. Example. Since Wa (v. Let A0 = − dx2 . limit point alternative We call τ limit circle (l. Weyl’s limit circle.) at a. . If τ is regular at a. ˆ Lemma 9. show that there is a function f ∈ D(τ ) restricted to [c.21). . 1)qf ∈ L2 (0. (pf )(c) = β and f (d) = γ. selfadjointness seems to be related to the Wronskian 0 of two functions at the boundary. Similarly for b. γ. f ) = 0 for at least one f ∈ D(τ ). f3 = ˆ ˆ v ∗ .22) ⇔ Wa (v. h) = Wa (f. (pv )(a)) = (0.23) which remains valid in the limit x → a. f ) = (pf )(a)v(a) − (pv )(a)f (a) any realvalued v with (v(a). Weyl’s limit circle. (pf )(c) = δ.24) . then f ∈ D(A0 ) if and only if f (a) = (pf )(a) = 0. Moreover. h) + Wb (f. f ∗ ) = 0 ∀f ∈ D(τ ) (9. Suppose v ∈ D(τ ) with Wa (v ∗ . . g) = 0 ⇒ Wa (g ∗ . Given α. and B = q. f2 = f. Find a q ∈ L1 (0.2.) at a if there is a v ∈ D(τ ) with Wa (v ∗ . (Hint: Problem 0. f2 = g ∗ . h) = 0 shows f ∈ D.1.c. g ∈ D(τ ) (9. h) = 0. f4 = f we infer (9. 1)}. b) such that f (c) = α. v) = 0 and there is a f ∈ D(τ ) ∗ . f2 )Wx (f3 . v) = 0 such that Wa (v. f ) = 0 ∀f. 1]f (0) = f (1) = 0}. This result shows that any selfadjoint extension of A0 must lie between A0 and A∗ . f2 ) + Wx (f1 . Choosing f1 = v.1). If τ is regular at a.5. For all f1 . 0) works. d] ⊆ (a. Otherwise τ is called limit point (l. h) = −Wa (f. δ.9. f4 )Wx (f2 . 1) such that D(A0 ) ∩ D(B) = {0}. β. f3 )Wx (f4 .2. f ) = 0.
ua (z. Example. x) which is in L2 ((a. y)g(y)r(y)dy. f ) = 0 if l. g) = 0 for all f ∈ D(A). Theorem 9. f )a = 0 for some f ∈ D(τ ). (9. Similarly. Proof. v) = 0 is trivially satisﬁed in this case. π). z ∈ C\R which satisfy Wc (u∗ . Similarly.28) where G(z. (9. f ) = cos(α)f (a) + sin(α)(pf )(a). at a.6. Lemma 9. Let g ∈ D(A∗ ).c. we can choose f such that it coincides with v near a and hence Wa (v.25) The most common choice α = 0 is known as Dirichlet boundary condition f (a) = 0.27) (9. Moreover. g) = 0 for all f. y) = 1 W (ub (z). ua (z)) ub (z.c.7.162 9. if τ is l. Then the operator A : D(A) → L2 (I. such that Wa (v. In particular. then Wa (f.c. As in the computation of A0 0 we conclude Wa (f. g) = 0.c. then let v ∈ D(τ ) with W (v ∗ . at b} is selfadjoint. Clearly A ⊆ A∗ ⊆ A∗ . v)a = 0 and W (v. y) x ≤ y (9. The resolvent of A is given by (A − z)−1 g(x) = a b G(z. then there exists a solution ua (z. α ∈ [0. x. at a. there exists a solution ub (z. (pv )(a)) = (sin(α). let w be an analogous function. who considered the set of solutions τ u = zu. g ∈ D(τ ). y) x ≥ y . f ) = 0 if l. One dimensional Schr¨dinger operators o Note that if Wa (f. x)ua (z. v) = 0. at a Wb (w. c). that is g ∈ D(A). Next we want to compute the resolvent of A.c. x)ub (z. The name limit circle respectively limit point stems from the original approach of Weyl. x) with the analogous properties near b. Hence it is no restriction to assume that v is real and Wa (v ∗ . − cos(α)). r dx) and which satisﬁes the boundary condition at a if τ is l. If τ is l. at b. u) = 0. τ is limit point if and only if Wa (f. Re(v)) = 0 or Wa (f. r dx) f → τf with D(A) = {f ∈ D(τ ) Wa (v. x.29) . Im(v)) = 0. They can be shown to lie on a circle which converges to a circle respectively point as c → a (or c → b). If τ is regular at a we can choose v such that (v(a). Suppose z ∈ ρ(A).26) (9.
say ub (z). (9. we can choose ua (z.30) ˜ Near a (x < c) we have f (x) = αu1 (x) + βu2 (x) and near b (x > d) we have b b ˜ f (x) = αu1 (x) + βu2 (x). x) . Similarly we obtain ub . r dx) be realvalued and consider f = (A − z)−1 g ∈ c D(A).c. Weyl’s limit circle. x) to be the solution corresponding to the initial conditions (ua (z. ua (z. r dx). The only problem is that ua or ub might be identically zero. If τ is regular at both a and b there is a corresponding solution ub (z. a). x) exists for all z ∈ C.2. x. f ) = Wb (ub . Since this set is dense the claim follows. we must have β = 0 since βu2 = f − αub is. a)) = (sin(α). So the only values of z for which (A − z)−1 does not exist must be those with W (ub (z). However. x) by setting it equal to f near a and using the diﬀerential equation to extend it to the rest of I. u2 ) = 1 we see that τ is l. again for all z. ˜ ˜ ˜ If f vanishes identically near both a and b we must have α = β = α = β = 0 ˜ b and thus α = β = 0 and a uj (y)g(y)r(y)dy = 0.32) = a G(z. αub + βu2 ) = β. If u2 is ˜ square integrable. (9. ua (z)) = 0. 2. Choosing u2 = ua and arguing as before we see α = 0 and hence x b f (x) = ub (x) a b ua (y)g(y)r(y)dy + ua (x) x ub (y)g(y)r(y)dy (9. Now choose u1 = ub and consider the behavior near b. (pua )(z. In particular. x). Since (τ − z)f = 0 near a respectively b. Let g ∈ L2 (I. Hence we need to show that this can be avoided by choosing g properly. we obtain ua (z. d) ⊂ I. If u2 is not square integrable on (d. where α = α + a u2 gr dy and β = β + a u1 gr dy.9. c Example. y)g(y)r(y)dy for any g ∈ L2 (I. Since (τ − z)f = g we have x b f (x) = u1 (x) α + a u2 gr dy + u2 (x) β + x u1 gr dy . − cos(α)). Since W (ub . Thus β = 0 in both cases and we ˜ have x b f (x) = ub (x) α + a b u2 gr dy + u2 (x) x ub gr dy. at a and hence 0 = Wb (ub . If τ is regular at a with a boundary condition as in the previous example. we can ﬁnd two functions in D(τ ) which coincide with ub and u2 near b. in this case ua (z. This case can be avoided choosing g suitable and hence there is at least one solution. Fix z and let g be supported in (c. limit point alternative 163 Proof.31) Now choosing g such that a ub gr dy = 0 we infer existence of ua (z). b). j = 1.
164
9. One dimensional Schr¨dinger operators o
and ub (z, x) are linearly dependent and ua (z, x) = γub (z, x) satisﬁes both boundary conditions. That is, z is an eigenvalue in this case. In particular, regular operators have pure point spectrum. We will see in Theorem 9.10 below, that this holds for any operator which is l.c. at both end points. In the previous example ua (z, x) is holomorphic with respect to z and satisﬁes ua (z, x)∗ = ua (z ∗ , x) (since it coresponds to real initial conditions and our diﬀerential equation has real coeﬃcients). In general we have: Lemma 9.8. Suppose z ∈ ρ(A), then ua (z, x) from the previous lemma can be chosen locally holomorphic with respect to z such that ua (z, x)∗ = ua (z ∗ , x). Similarly, for ub (z, x). Proof. Since this is a local property near a we can suppose b is regular and choose ub (z, x) such that (ub (z, b), (pub )(z, b)) = (sin(β), − cos(β)) as in the example above. In addition, choose a second solution vb (z, x) such that (vb (z, b), (pvb )(z, b)) = (cos(β), sin(β)) and observe W (ub (z), vb (z)) = 1. If z ∈ ρ(A), z is no eigenvalue and hence ua (z, x) cannot be a multiple of ub (z, x). Thus we can set ua (z, x) = vb (z, x) + m(z)ub (z, x) (9.34) and it remains to show that m(z) is holomorphic with m(z)∗ = m(z ∗ ). Choosing h with compact support in (a, c) and g with support in (c, b) we have h, (A − z)−1 g = h, ua (z) g ∗ , ub (z) (9.35) = ( h, vb (z) + m(z) h, ub (z) ) g ∗ , ub (z) (9.33)
(with a slight abuse of notation since ub , vb might not be square integrable). Choosing (realvalued) functions h and g such that h, ub (z) g ∗ , ub (z) = 0 we can solve for m(z): m(z) = This ﬁnishes the proof.
d Example. We already know that τ = − dx2 on I = (−∞, ∞) gives rise to the free Schr¨dinger operator H0 . Furthermore, o √
2
h, (A − z)−1 g − h, vb (z) g ∗ , ub (z) . h, ub (z) g ∗ , ub (z)
(9.36)
(9.37) √ are two linearly independent solutions (for z = 0) and since Re( −z) > 0 for z ∈ C\[0, ∞), there is precisely one solution (up to a constant multiple)
u± (z, x) = e
−zx
,
z ∈ C,
9.2. Weyl’s limit circle, limit point alternative
165
which is square integrable near ±∞, namely u± . In particular, the only choice for ua is u− and for ub is u+ and we get
√ 1 G(z, x, y) = √ e− −zx−y 2 −z
(9.38)
which we already found in Section 7.4. If, as in the previous example, there is only one square integrable solution, there is no choice for G(z, x, y). But since diﬀerent boundary conditions must give rise to diﬀerent resolvents, there is no room for boundary conditions in this case. This indicates a connection between our l.c., l.p. distinction and square integrability of solutions. Theorem 9.9 (Weyl alternative). The operator τ is l.c. at a if and only if for one z0 ∈ C all solutions of (τ − z0 )u = 0 are square integrable near a. This then holds for all z ∈ C. Similarly for b. Proof. If all solutions are square integrable near a, τ is l.c. at a since the Wronskian of two linearly independent solutions does not vanish. Conversely, take two functions v, v ∈ D(τ ) with Wa (v, v ) = 0. By con˜ ˜ sidering real and imaginary parts it is no restriction th assume that v and v are realvalued. Thus they give rise to two diﬀerent selfadjoint operators ˜ ˜ A and A (choose any ﬁxed w for the other endpoint). Let ua and ua be the ˜ corresponding solutions from above, then W (ua , ua ) = 0 (since otherwise ˜ ˜ A = A by Lemma 9.5) and thus there are two linearly independent solutions which are square integrable near a. Since any other solution can be written as a linear combination of those two, every solution is square integrable near a. It remains to show that all solutions of (τ − z)u = 0 for all z ∈ C are square integrable near a if τ is l.c. at a. In fact, the above argument ensures ˜ this for every z ∈ ρ(A) ∩ ρ(A), that is, at least for all z ∈ C\R. Suppose (τ − z)u = 0. and choose two linearly independent solutions uj , j = 1, 2, of (τ − z0 )u = 0 with W (u1 , u2 ) = 1. Using (τ − z0 )u = (z − z0 )u and (9.10) we have (a < c < x < b)
x
u(x) = αu1 (x) + βu2 (x) + (z − z0 )
c
(u1 (x)u2 (y) − u1 (y)u2 (x))u(y)r(y) dy. (9.39)
Since uj ∈
L2 ((c, b), rdx)
we can ﬁnd a constant M ≥ 0 such that
b
u1,2 (y)2 r(y) dy ≤ M.
c
(9.40)
166
9. One dimensional Schr¨dinger operators o
Now choose c close to b such that z − z0 M 2 ≤ 1/4. Next, estimating the integral using Cauchy–Schwarz gives
x
(u1 (x)u2 (y) − u1 (y)u2 (x))u(y)r(y) dy
c x
2
x
≤
c
u1 (x)u2 (y) − u1 (y)u2 (x)2 r(y) dy
c x
u(y)2 r(y) dy (9.41)
x
≤ M u1 (x)2 + u2 (x)2
c
u(y)2 r(y) dy
and hence
x
u(y)2 r(y) dy ≤ (α2 + β2 )M + 2z − z0 M 2
c c
u(y)2 r(y) dy (9.42)
≤ (α2 + β2 )M + Thus
c x
1 2
x
u(y)2 r(y) dy.
c
u(y)2 r(y) dy ≤ 2(α2 + β2 )M
(9.43)
and since u ∈ ACloc (I) we have u ∈ L2 ((c, b), r dx) for every c ∈ (a, b). Note that all eigenvalues are simple. If τ is l.p. at one endpoint this is clear, since there is at most one solution of (τ − λ)u = 0 which is square integrable near this end point. If τ is l.c. this also follows since the fact that two solutions of (τ − λ)u = 0 satisfy the same boundary condition implies that their Wronskian vanishes. Finally, led us shed some additional light on the number of possible boundary conditions. Suppose τ is l.c. at a and let u1 , u2 be two solutions of τ u = 0 with W (u1 , u2 ) = 1. Abbreviate
j BCx (f ) = Wx (uj , f ),
f ∈ D(τ ).
(9.44) (9.45)
Let v be as in Theorem 9.6, then, using Lemma 9.5 it is not hard to see that Wa (v, f ) = 0 ⇔
1 2 cos(α)BCa (f ) + sin(α)BCa (f ) = 0,
1 BCa (v) 2 BCa (v)
where tan(α) = − . Hence all possible boundary conditions can be parametrized by α ∈ [0, π). If τ is regular at a and if we choose u1 (a) = p(a)u2 (a) = 1 and p(a)u1 (a) = u2 (a) = 0, then
1 BCa (f ) = f (a), 2 BCa (f ) = p(a)f (a)
(9.46) (9.47)
and the boundary condition takes the simple form cos(α)f (a) + sin(α)p(a)f (a) = 0. Finally, note that if τ is l.c. at both a and b, then Theorem 9.6 does not give all possible selfadjoint extensions. For example, one could also choose
1 1 BCa (f ) = eiα BCb (f ), 2 2 BCa (f ) = eiα BCb (f ).
(9.48)
9.2. Weyl’s limit circle, limit point alternative
167
The case α = 0 gives rise to periodic boundary conditions in the regular case. Now we turn to the investigation of the spectrum of A. If τ is l.c. at both endpoints, then the spectrum of A is very simple Theorem 9.10. If τ is l.c. at both end points, then the resolvent is a Hilbert– Schmidt operator, that is,
b a a b
G(z, x, y)2 r(y)dy r(x)dx < ∞.
(9.49)
In particular, the spectrum of any self adjoint extensions is purely discrete and the eigenfunctions (which are simple) form an orthonormal basis. Proof. This follows from the estimate
b a a b b x b
ub (x)ua (y)2 r(y)dy +
x
ub (y)ua (x)2 r(y)dy r(x)dx (9.50)
≤2
a
ua (y)2 r(y)dy
a
ub (y)2 r(y)dy,
which shows that the resolvent is Hilbert–Schmidt and hence compact. If τ is not l.c. the situation is more complicated and we can only say something about the essential spectrum. Theorem 9.11. All self adjoint extensions have the same essential spectrum. Moreover, if Aac and Acb are selfadjoint extensions of τ restricted to (a, c) and (c, b) (for any c ∈ I), then σess (A) = σess (Aac ) ∪ σess (Acb ). (9.51)
Proof. Since (τ − i)u = 0 has two linearly independent solutions, the defect indices are at most two (they are zero if τ is l.p. at both end points, one if τ is l.c. at one and l.p. at the other end point, and two if τ is l.c. at both endpoints). Hence the ﬁrst claim follows from Theorem 6.19. For the second claim restrict τ to the functions with compact support in (a, c) ∪ (c, d). Then, this operator is the orthogonal sum of the operators A0,ac and A0,cb . Hence the same is true for the adjoints and hence the defect indices of A0,ac ⊕ A0,cb are at most four. Now note that A and Aac ⊕ Acb are both selfadjoint extensions of this operator. Thus the second claim also follows from Theorem 6.19.
d Problem 9.3. Compute the spectrum and the resolvent of τ = − dx2 , I = (0, ∞) deﬁned on D(A) = {f ∈ D(τ )f (0) = 0}.
2
168
9. One dimensional Schr¨dinger operators o
9.3. Spectral transformations
In this section we want to provide some fundamental tools for investigating the spectra of SturmLiouville operators and, at the same time, give some nice illustrations of the spectral theorem.
d Example. Consider again τ = − dx2 on I = (−∞, ∞). From Section 7.2 we know that the Fourier transform maps the associated operator H0 to the multiplication operator with p2 in √2 (R). To get multiplication by λ, as in L the spectral theorem, we set p = λ and split the Fourier integral into a positive and negative part
2
(U f )(λ) = Then
i λx R e √ f (x) dx −i λx f (x) dx Re 2
√
,
λ ∈ σ(H0 ) = [0, ∞).
(9.52)
U : L (R) →
j=1
2
L2 (R,
χ[0,∞) (λ) √ dλ) 2 λ
(9.53)
is the spectral transformation whose existence is guaranteed by the spectral theorem (Lemma 3.3). Note that in the previous example the kernel e±i λx of the integral transform U is just a pair of linearly independent solutions of the underlying diﬀerential equation (though no eigenfunctions, since they are not square integrable). More general, if U : L2 (I, r dx) → L2 (R, dµ), f (x) →
R √
u(λ, x)f (x)r(x) dx
(9.54)
is an integral transformation which maps a selfadjoint SturmLiouville operator A to multiplication by λ, then its kernel u(λ, x) is a solution of the underlying diﬀerential equation. This formally follows from U Af = λU f which implies u(λ, x)(τ − λ)f (x)r(x) dx =
R R
(τ − λ)u(λ, x)f (x)r(x) dx
(9.55)
and hence (τ − λ)u(λ, .) = 0. Lemma 9.12. Suppose
k
U : L2 (I, r dx) →
j=1
L2 (R, dµj )
(9.56)
is a spectral mapping as in Lemma 3.3. Then U is of the form
b
U f (x) =
a
u(λ, x)f (x)r(x) dx,
(9.57)
) is a solution of τ uj = λuj for a.58) Moreover. d]. Next. uk (λ. b).c. uj (λ. . . since we have Uj χ[c.d] (λ.62) Interchanging integrals on the right hand side (which is permitted at least for g. the formula for the inverse follows. x).d] = Uj χ[ˆ.9 there is a corresponding kernel uj b [c. if τ is l.d] . b). x)) and each uj (λ. (9.d] f )(λ) = a uj [c. b). λ (with respect to µj ). x.61) for every f with compact support in (a. . they satisfy the boundary condition. d]. b) × (a.d] = (λ − z)Uj RA (z)χ[c. d] ⊂ (a.) = uj (λ. . F with compact support). Spectral transformations 169 where u(λ.d] . for ordered spectral measures we have always k ≤ 2 and even k = 1 if τ is l. Using Uj RA (z) = 1 λ−z Uj we have b Uj f (x) = (λ − z)Uj a G(z.c. y)f (y)r(y) dy. (9.d] is Hilbert–Schmidt since G(z. In particular.d] χ[c. we see Fj (λ)∗ j R a b b uj (λ. this formula holds for any f (provided the integral is understood as the corresponding limit). at some endpoint. U g = U −1 F. the solutions uj (λ) are linearly independent if the spectral measures are ordered and.d] c ˆ Uj f (x) = a uj (λ. The inverse is given by k U −1 F (λ) = j=1 R uj (λ. F. then the kernels coincide c ˆ on [c. Proof.e. there is a kernel uj (λ.d] [ˆ.d] (y) is square integrable over (a.d] (λ. x. . then RA (z)χ[c. x)f (x)r(x) dx (9. Since functions with compact support are dense and Uj is continuous.9. Now take a larger compact interval [ˆ. from Uj Af = λUj f we have b b uj (λ.59) If we restrict RA (z) to a compact interval [c. y) such that (9. Hence Uj χ[c. y)χ[c. d] ⊇ [c. x)f (x)r(x) dx (9. (9. Using the fact that U is unitary. at one endpoint. .60) (Uj χ[c.3. x)(τ f )(x)r(x) dx = λ a a uj (λ.d] is Hilbert–Schmidt as well and by Lemma 6. x)∗ Fj (λ)dµj (λ). x) such that b [c.63) . c ˆ In particular. . x) = (u1 (λ. g .)χ[c. x)f (x)r(x) dx. x)g(x)r(x) dx = a (U −1 F )(x)∗ g(x)r(x) dx.
x) = γa (λ)s(λ. x) is the solution which satisﬁes the boundary condition at a.64) then we have l cj (λ)Fj (λ) = 0. are linearly independent for λ ∈ Ωl which shows k ≤ 2 since there are at most two linearly independent solutions. . Restricting everything to [c. For simplicity we will only pursue the case where one endpoint. . a) = sin(α). (9. Similarly for (9. x) = 0 j=1 (9. λ and every f ∈ D(A0 ).57) has to be understood as β Uj f (x) = lim α↓a. Please note that the integral in (9. if uj (λ. Fj (λ) = 1 for j = j0 and Fj (λ) = 0 else which shows cj0 (λ) = 0.67) and introduce two solution s(z.11.e. .0 ab. a) = cos(α). d] ⊂ (a. In particular.0 λuj (λ. ﬁx l ≤ k. If we assume the µj are ordered. dµj ). If τ is l. Suppose l cj (λ)uj (λ. Finally. we can choose ua (z. Moreover. that is. b) and splitting A as in Theorem 9. there is a set Ωl such that µj (Ωl ) = 0 for 1 ≤ j ≤ l. near a. say a. and uj (λ. e.g..c.β↑b α uj (λ. x) and using the rescaling dµ(λ) = .170 9. The general case can usually be reduced to this case by choosing c ∈ (a. Moreover. b) the above equation implies uj (λ.66) where the limit is taken in L2 (R.58). One dimensional Schr¨dinger operators o for a.65) for every f . j=1 Fj = Uj f.d] ∈ D(A∗ ) and A∗ uj (λ.) is τ is l.d] = cd. x) must satisfy the boundary condition. We choose a boundary condition cos(α)f (a) + sin(α)p(a)f (a) = 0 (9. Hence uj (λ. x)f (x)r(x) dx.) is a solution of τ uj = λuj .)[c. . x) and c(z. there is only one linearly independent solution and thus k = 1. x).c. a) = cos(α). x) = s(z.68) Note that s(z. (9. in our previous lemma we have u1 (λ.)[c. . Since U is surjective. 1 ≤ j ≤ l. uj (λ. (9. c(z. we can prescribe Fj arbitrarily. p(a)c (z. we can choose α = a and f to satisfy the boundary condition. p(a)s (z. a) = − sin(α).)[c. x) of τ u = zu satisfying the initial conditions s(z.d] . x). is regular.
75) (9. the above formula holds at least for the pure point part µpp . Furthermore. Example. r dx) → L2 (R. For arbitrary A. with inverse (U −1 F )(x) = (U f )(λ) = a s(λ. cos(α)ub (z.70) Note however. R (9. For µ1 we know µ1 (R) = 1. So our next aim must be to compute µ. x)). x) and s(z. the measure µ is the object of desire since it contains all the spectral information of A. a) + sin(α)p(a)ub (z. it turns out that µ is indeed unbounded.72) that is µ({E}) = s(E) In particular. only eigenvalues) this is straightforward as the following example shows.74) .e.69) s(λ. it destroys the normalization of the measure µ. x)F (λ)dµ(λ). a) z ∈ ρ(A). (9. λ = E . x) is the corresponding eigenfunction and hence the same is true for SE (λ) = (U s(E))(λ). we have ∞ dµ(λ) = j=1 1 s(Ej ) 2 dΘ(λ − Ej ). (9. In particular. dµ). x) are linearly independent solutions. In the general case we have to work a bit harder. a) − sin(α)ub (z. j=1 (9. x)2 r(x)dx = R −2 . Then s(E. that while this rescaling gets rid of the unknown factor γa (λ).3. we can write ub (z. So up to this point we have our spectral transformation U which maps A to multiplication by λ. 0.. Spectral transformations 171 γa (λ)2 dµa (λ) and (U1 f )(λ) = γa (λ)(U f )(λ) we obtain a unitary map b U : L2 (I. SE (λ)2 dµ(λ) = s(E) 4 µ({E}). where mb (z) = cos(α)p(a)ub (z. but µ might not even be bounded! In fact. x) + mb (z)s(z. since U is unitary we have b s(E) 2 = a s(E.9.73) where dΘ is the Dirac measure centered at 0. Since c(z..71) Moreover. x) = γb (z)(c(z. σp (A) = {Ej }∞ . If A has only pure point spectrum (i. that is SE (λ) = s(E) 2 . W (c(z). Suppose E ∈ σp (A) is an eigenvalue. if τ is limit circle at both endpoints). a) . λ = E. SE (λ) = 0 for a. s(z)) = 1.g.e. if A has pure point spectrum (e. x)f (x)r(x)dx (9. λ=0 (9. but we know nothing about the measure µ.
x)F (λ) dµ(λ). the right hand side is continuous with respect to x and so is the left hand side.13.77). Here equality is to be understood in L2 . First of all note that from RA (z)f = U −1 λ−z U f we have b 1 . x) (the denominator in (9. x) is normalized as in (9. x) satisﬁes the boundary condition at a. x − 2Im(z) a ub (z. x. x) and v(x) = ub (z. now diﬀerentiate with respect to x). x). y)f (y)r(y) dy = sin(α) R F (λ) dµ(λ).83) .77) Lemma 9. Hence in this case the formula holds for all x and we can choose x = a to obtain b sin(α) a ub (z.79) (clearly it is true for x = a. y)2 r(y) dy = Wx (ub (z). (9. λ−z (9. λ−z (9. v(x) of τ u = zu.76) since the same is true for ub (z. We have (U ub (z))(λ) = where ub (z. at least if F has compact support. The Weyl mfunction is a Herglotz function and satisﬁes b Im(mb (z)) = Im(z) a ub (z. if z is an eigenvalue). (9. Moreover. Lemma 9. since both ub and u∗ are in b b D(τ ) near b. y)f (y)r(y) dy = a R s(λ. However.80) and observe that Wx (ub . v) − Wa (u. u∗ ) vanishes as x ↑ b. x) = c(z.14. One dimensional Schr¨dinger operators o is known as WeylTitchmarsh mfunction. λ−z (9.78) where ub (z. (9. x)2 r(x) dx. 1 Proof. τ v = z v it is straightforˆ ward to check x (ˆ − z) z a u(y)v(y)r(y) dy = Wx (u. Now choose u(x) = ub (z.e. v) (9. Proof. x) + mb (z)s(z.77). x)∗ = ub (z ∗ .81) G(z. Note that mb (z) is holomorphic in ρ(A) and that mb (z)∗ = mb (z ∗ ) (9. that is for a. ub (z. Given two solutions u(x).82) where F = U f .172 9. x). x) is normalized as in (9. the constant γb (z) is of no importance and can be chosen equal to one. that is. x. ub (z)∗ ) − 2Im(mb (z)).75) only vanishes if ub (z.
Spectral transformations 173 for all f .85) and d = Re(mb (i)). 0)e− and thus √ −zx (9.3. x) = cos(α) cos( zx) + √ sin( zx) z and √ √ cos(α) s(z. where F has compact support. Stieltjes inversion formula follows as in the case where the measure is bounded. Consider τ = − dx2 on I = (0.82) with respect to x before setting x = a.91) (9. α = 0. The Weyl mfunction is given by mb (z) = d + R 1 λ − λ − z 1 + λ2 dµ(λ). (9. that is.87) where Im(mb (λ + iε)) = ε a ub (λ + iε.86) and hence the right hand side 1 λ of (9.84) λ − z2 and since a holomorphic function is determined up to a real constant by its imaginary part we obtain Theorem 9. Since these functions are dense.86) Moreover. x)2 r(x) dx. ﬁrst diﬀerentiate (9. x)2 r(x) dx = Im(z) R 1 dµ(λ) (9. the claim follows if we can cancel sin(α). Now combining the last two lemmas we infer from unitarity of U that b Im(mb (z)) = Im(z) a ub (z. To see the case α = 0.92) √ −z cos(α) + sin(α) mb (z) = √ −z sin(α) − cos(α) . Then 2 Im(z) λ−z2 √ √ sin(α) c(z.88) Proof. µ is given by Stieltjes inversion formula µ(λ) = lim lim δ↓0 ε↓0 Im(mb (λ + iε))dλ. x) = ub (z. R 1 dµ(λ) = Im(mb (i)) < ∞. ∞).15. (9. d Example. 1 + λ2 1 π λ+δ (9.9. (9. By Im( λ−z − 1+λ2 ) = its imaginary part coincides with that of mb (z) and hence equality follows.90) Moreover. z ub (z. Choosing z = i in (9.84) shows (9. d ∈ R.89) (9.85) is a welldeﬁned holomorphic function in C\R. δ b (9. x) = − sin(α) cos( zx) + √ sin( zx).
(9. x. x. y) − Gk (z.99) (9. Furthermore we have Lemma 9. Proof.98) .94) R λ−z in this case (the factor − cot(α) follows by considering the limit z → ∞ of both sides). y).93) Note that if α = 0 we even have and hence < 0 in the previous example 1 dµ(λ) (9. We ﬁrst observe (Problem 9. x.96) (9. since we know equality only for a. The Green’s function of A G(z. RA (z)ϕk (. π(cos(α)2 + λ sin(α)2 ) 1 λ−z dµ(λ) √ (9. For simplicity we only consider the case p = k ≡ 1. a more careful analysis is needed. y) k ϕ 2 + G(z. x. (9. One dimensional Schr¨dinger operators o respectively dµ(λ) = λ dλ. y) + G(z. One can show that this remains true in the general case. x) = cos(α) + √ e (1 + O( √ )) 2 −z −z √ 1 cos(α) 1 s(z.97) G(z. Then. 1 Formally it follows by choosing x = a in ub (z. x) = (U −1 λ−z )(x). x. x) = − sin(α) + √ e −z(x−a) (1 + O( √ )) (9.e.17. y) = s(z. b). y) = ϕk (. y) − Gk (z.4): Lemma 9.14). x. y) = 0 for every x. by RA (z) ≤ Im(z)−1 (see Theorem 2. y) ≤ Gk (z. y ∈ (a. x)ub (z. − y) . G(z. however. Let ϕ be some continuous positive function supported in [−1. x. 1] with ϕ(x)dx = 1 and set ϕk (x) = k ϕ(k x). satisﬁes lim Im(z)→∞ x ≤ y. ≤ Im(z) where Gk (z. mb (z) = cot(α) + We conclude this section with some asymptotics for large z away from the spectrum. − x). x.174 9. b) as Im(z) → ∞.16. y). x. We have 1 sin(α) √−z(x−a) 1 c(z.95) 2 −z −z for ﬁxed x ∈ (a. x.
x. b)) one can obtain further asymptotic terms in Lemma 9.101) s(z. Prove the asymptotics for c(z.2 and consider c(z. Note that assuming q ∈ C k ([a. Problem 9. x) and s(z. y) ≤ ε an the claim follows.100) . x. x. y) ≤ 2 and then Im(z) so large such that k ϕ 2 Im(z) ε ≤ 2 .) cot(α) + O( √1 ). The Weyl mfunction satisﬁes mb (z) = as Im(z) → ∞. y)dy −z a by Lemma 9.9. x. Combining the last two lemmas we also obtain Lemma 9.16 into Lemma 9. −z α=0 α=0 (9.16. x) mb (z) = − + o(e−2 −z(x−a) ) (9. y) we can ﬁrst choose k suﬃciently large ε such that G(z. x) from Lemma 9. x) = cos(α) cosh( −zx) + √ sinh( −zx) −z x √ 1 −√ sinh( −z(x − y))q(y)c(z. Hence G(z. (Hint: Use that √ √ sin(α) c(z. Proof.17 we see √ c(z.3. x) form which the claim follows.4. x) = e− ˜ √ −zx c(z.18. y) − Gk (z.16 and hence also in the expansion of mb (z). −z √ − −z + O( √1 ). Spectral transformations 175 Since G(z. Plugging the asymptotic expansions from Lemma 9. y) = limk→∞ Gk (z. x). x.
.
The Hamilo tonian of one particle in d dimensions is given by H = H0 + V. since (p2 + γ 2 )ψ ∈ L2 (Rn ).Chapter 10 Oneparticle Schr¨dinger operators o 10. Lemma 10. (10.2) Proof. (10. The important observation is that (p2 + γ 2 )−1 ∈ L2 (Rn ) if n ≤ 3. (10. We are mainly interested in the case 1 ≤ d ≤ 3 and want to ﬁnd classes of potentials which are relatively bounded respectively relatively compact. Selfadjointness and spectrum Our next goal is to apply these results to Schr¨dinger operators. To do this we need a better understanding of the functions in the domain of H0 .4) 177 . Then ψ ∈ C∞ (Rn ) and for any a > 0 there is a b > 0 such that ψ ∞ ≤ a H0 ψ + b ψ . But now everything follows from the RiemannLebesgue lemma ψ ∞ ˆ ˆ ≤ (2π)−n/2 (p2 + γ 2 )−1 ( p2 ψ(p) + γ 2 ψ(p) ) = (γ/2π)n/2 (p2 + 1)−1 (γ −2 H0 ψ + ψ ) (10. ˆ Hence. Suppose n ≤ 3 and ψ ∈ H 2 (Rn ).3) ˆ shows ψ ∈ L1 (Rn ).1.1) where V : Rd → R is the potential energy of the particle. the CauchySchwarz inequality ˆ ψ 1 = ≤ ˆ (p2 + γ 2 )−1 (p2 + γ 2 )ψ(p) 1 ˆ (p2 + γ 2 )−1 (p2 + γ 2 )ψ(p) .1.
(10.178 10.8) and that H (1) is bounded from below E0 = inf σ(H (1) ) > −∞. D(H (1) ) = H 2 (R3 ). ∞) (10. Our previous lemma shows D(H0 ) ⊆ D(V ) and the rest follows from Lemma 7. Then V is relatively compact with respect to H0 .2. ∞ In particular. (10. Theorem 10. . we might have E0 < 0 and there might be some discrete eigenvalues below the essential spectrum. Now we come to our ﬁrst result. if γ > 0. but if γ > 0.9) If γ ≤ 0 we have H (1) ≥ 0 and hence E0 = 0. Note that f ∈ L∞ (Rn ) ∩ L2 (Rn ) for n ≤ 3.7) x If the potential is attracting. ∞ Moreover. (10. D(H) = H 2 (Rn ). ∞). Theorem 10. Let V be realvalued and V ∈ L∞ (Rn ) if n > 3 and V ∈ ∞ L∞ (Rn ) + L2 (Rn ) if n ≤ 3.10 using f (p) = (p2 − z)−1 and g(x) = V (x). Oneparticle Schr¨dinger operators o ﬁnishes the proof. 1 As domain we have chosen D(H (1) ) = D(H0 ) ∩ D( x ) = D(H0 ) and by Theorem 10. C0 (Rn ) is a core for H.5) is selfadjoint.2 we conclude that H (1) is selfadjoint. we must have V ∈ L2 (Rn ) if loc D(V ) ⊆ D(H0 ). The hydrogen atom We begin with the simple model of a single electron in R3 moving in the external potential V generated by a nucleus (which is assumed to be ﬁxed at the origin). 10. H = H0 + V. that is.2. (10. bounded from below and σess (H) = [0.6) Proof. then V is given by the Coulomb potential and the corresponding Hamiltonian is given by γ H (1) = −∆ − . ∞ ∞ Observe that since Cc (Rn ) ⊆ D(H0 ). Moreover.2 also tells us σess (H (1) ) = [0. then it describes the hydrogen atom and is probably the most famous model in quantum mechanics. If one takes only the electrostatic force into account.
(10. (10. (2H0 + V (1) )ψ . Theorem 10. This result even has some further consequences for the point spectrum of H (1) .13) D(H (1) (s)) = D(H (1) ). Since each eigenvalue Ej has ﬁnite multiplicity (it lies in the discrete spectrum) there must be an inﬁnite number of eigenvalues which accumulate at 0.10.11) 2 2 Now let us investigate the action of U (s) on H (1) H (1) (s) = U (−s)H (1) U (s) = e−2s H0 + e−s V (1) . Consider the dilation group U (s)ψ(x) = e−ns/2 ψ(e−s x). with limj→∞ Ej = 0. Then Theorem 4. Now choose a sequence sn → ∞ such that we have supp(ψ(sn )) ∩ supp(ψ(sm )) = ∅ for n = m. s ∈ R. Hψ − Hψ. Suppose γ > 0. H]ψ = U (−s)ψ. (10. Choose ψ ∈ Cc (R\{0}) and set ψ(s) = U (−s)ψ.16) ψ(s). U (s)ψ = 0 and hence 0 = lim 1 H − H(s) ψ. ψ s→0 s s→0 s = ψ. ψ ∈ S(Rn ). Corollary 10.15) 2 In particular. The generator can be easily computed 1 in Dψ(x) = (xp + px)ψ(x) = (xp − )ψ(x). Then any normalized eigenfunction ψ corresponding to an eigenvalue λ satisﬁes 1 λ = − ψ. [U (s).11 (i) shows that rank(PH (1) ((−∞. [U (s).3. (10.12) (10. . Then E0 < Ej < Ej+1 < 0.17) which is negative for s large. The hydrogen atom 179 In order to say more about the eigenvalues of H (1) we will use the fact that both H0 and V (1) = −γ/x have a simple behavior with respect to scaling. then ψ. 0))) = ∞. Then σp (H (1) ) = σd (H (1) ) = {Ej−1 }j∈N0 . V (1) ψ (10. H]ψ = lim U (−s)ψ. H (1) ψ(s) = e−2s ψ. V ψ . H0 ψ = ψ.10) which is a strongly continuous oneparameter unitary group.4. Now suppose Hψ = λψ.2. Suppose H = H0 + V with U (−s)V U (s) = e−s V .14) Thus we have proven the virial theorem. H0 ψ + e−s ψ. (10. ∞ Proof. (10. all eigenvalues must be negative.
sin(θ)dθ) ⊗ L2 ((0. ϕ) = R(r)Θ(θ)Φ(ϕ) we obtain the following three SturmLiouville equations 1 d 2d l(l + 1) r + + V (r) R(r) = λR(r) 2 dr r dr r2 1 d d m2 − sin(θ) + Θ(θ) = l(l + 1)Θ(θ) sin(θ) dθ dθ sin(θ) d2 − 2 Φ(ϕ) = m2 Φ(ϕ) dϕ − (10. 2π). Hence we have gotten a quite complete picture of the spectrum of H (1) . x3 = r cos(θ).20) In these new coordinates (r.18) For a general potential this is hopeless. (10. 2 ∂r r ∂r r γ V (r) = − . which is given by the partial diﬀerential equation − ∆ψ(x) − γ ψ(x) = λψ(x).2. ∞).) Making the product ansatz (separation of variables) ψ(r. r2 dr) ⊗ L2 ((0.22) 1 ∂ 2∂ 1 r + 2 L2 + V (r). (10.19) which correspond to a unitary transform L2 (R3 ) → L2 ((0. . ϕ) our operator reads H (1) = − where L2 = L2 + L2 + L2 = − 1 2 3 1 ∂ ∂ 1 ∂2 sin(θ) − . π). x3 ) → (r. First of all.180 10. x2 = r sin(θ) sin(ϕ).21) (Recall the angular momentum operators Lj from Section 8.23) (10. we could try to compute the eigenvalues of H (1) (in the case γ > 0) by solving the corresponding eigenvalue equation. r (10. θ. ϕ) x1 = r sin(θ) cos(ϕ). x (10.24) The form chosen for the constants l(l + 1) and m2 is for convenience later on. but in our case we can use the rotational symmetry of our operator to reduce our partial diﬀerential equation to ordinary ones. sin(θ) ∂θ ∂θ sin(θ)2 ∂ϕ2 (10. θ. θ. Next. it suggests itself to switch to spherical coordinates (x1 . dϕ). These equations will be investigated in the following sections. x2 . Oneparticle Schr¨dinger operators o If γ ≤ 0 we have σd (H (1) ) = ∅ since H (1) ≥ 0 in this case.
31) The operator τ transforms to the somewhat simpler form τm = − d d m2 (1 − x2 ) − . π).26) From our analysis in Section 9. (10.23) to be in the domain of H0 (in particular continuous). Φ(0) = Φ(2π). I = (0. dx dx 1 − x2 (10.28) Now we turn to the equation for Θ(θ) τm Θ(θ) = 1 sin(θ) − d d m2 sin(θ) + dθ dθ sin(θ) Θ(θ). 1). sin(θ)dθ) → L2 ((−1.3. 2π). m ∈ N0 .5. 2π). dx).29) (10. 2π form an orthonormal basis for L2 (0.25) is selfadjoint. Angular momentum We start by investigating the equation for Φ(ϕ) which associated with the StumLiouville equation τ Φ = −Φ .32) L2 ((0.30) For the investigation of the corresponding operator we use the unitary transform Θ(θ) → f (x) = Θ(arccos(x)). (10. π).1 we immediately obtain Theorem 10. . Its spectrum is purely discrete σ(A) = σd (A) = {m2 m ∈ Z} and the corresponding eigenfunctions 1 Φm (ϕ) = √ eimϕ . since in polar coordinates L3 = 1 ∂ .27) m ∈ Z. 2π].25) since we want ψ deﬁned via (10.3. we choose periodic boundary conditions the StumLiouville equation AΦ = τ Φ. Note that except for the lowest eigenvalue.10. The operator A deﬁned via (10. all eigenvalues are twice degenerate. Φ (0) = Φ (2π)} (10. (10. I = (0. 2π) Φ ∈ AC 1 [0. D(A) = {Φ ∈ L2 (0. . (10. i ∂ϕ (10. We note that this operator is essentially the square of the angular momentum in the third coordinate direction. Angular momentum 181 10.
34) dx where 1 dl 2 (x − 1)l (10. 1) it suﬃces to show that the functions (1 − x2 )−m/2 Plm (x) are dense. for m > 0 at both endpoints. Am is selfadjoint. note that Pl (x) are (nonzero) polynomials of degree l. (10. Now ﬁx l = 0 and note P0 (x) = 1. m ∈ N0 .6. linearly independent solution is given by x dt Qlm (x) = Plm (x) .37) Theorem 10. l ≥ m. For l ∈ N0 it is solved by the associated Legendre functions dm Plm (x) = (1 − x2 )m/2 m Pl (x).38) form an orthonormal basis for L2 (−1. Thus τm is l. For m > 0 we have Q0m = (x ± 1)−m/2 (C + O(x ± 1)) which shows that it is not square integrable. This is straightforward to check. To prove this. Moreover. By Theorem 9. Oneparticle Schr¨dinger operators o The corresponding eigenvalue equation τm u = l(l + 1)u (10. 1) .35) Pl (x) = l 2 l! dxl are the Legendre polynomials. for m = 0 and l. Its spectrum is purely discrete σ(Am ) = σd (Am ) = {l(l + 1)l ∈ N0 .p. (10. limx→±1 (1 − x2 )f (x) = 0} (10. π). it suﬃces to show that the functions Plm (x) are dense.c. Moreover.37) is selfadjoint. For m = 0 we have Q00 = arctanh(x) ∈ L2 and so τ0 is l. Since (1 − x2 ) > 0 for x ∈ (−1.23) is continuous) we choose the boundary condition generated by P0 (x) = 1 in this case x Am f = τ f.39) (10. for every SturmLiouville equation v(x) = u(x) satisﬁes p(t)u(t)2 τ v = 0 whenever τ u = 0.c.6. at both end points. (10. Proof. l ≥ m} and the corresponding eigenfunctions ulm (x) = 2l + 1 (l + m)! Plm (x). D(Am ) = {f ∈ L2 (−1. But the span of these functions . Plm is an eigenfunction corresponding to the eigenvalue l(l + 1) and it suﬃces to show that Plm form a basis.182 10.33) is the associated Legendre equation. 2 (l − m)! l ∈ N0 . In order to make sure that the eigenfunctions for m = 0 are continuous (such that ψ deﬁned via (10. 1). The operator Am . A second. τ f ∈ L2 (−1.36) 2 )P (t)2 lm 0 (1 − t dt In fact. 1) f ∈ AC 1 (0. deﬁned via (10.
. Ylm are smooth away from the origin and by construction they satisfy − ∆Ylm = l(l + 1) Ylm . π).43) ˜ where Plm is a polynomial of degree l − m given by dl+m ˜ Plm (x) = (1 − x2 )−m/2 Plm (x) = l+m (1 − x2 )l . 2 (l − m)! l = m. The only thing remaining is the normalization of the eigenfunctions.±m (x) = 2l + 1 (l + m)! ˜ x3 Plm ( ) 4π (l − m)! r x1 ± ix2 r m . π).40) form an orthonormal basis for L2 ((0.45) . dx (10. 4π (l − m)! lm m ≤ l. The spherical harmonics Ylm (θ.44) In particular. r = x. Note that transforming Ylm back to cartesian coordinates gives Yl. sin(θ)dθ) for any ﬁxed m ∈ N0 .10. Theorem 10.7. Everything follows from our construction. Returning to our original setting we conclude that Θlm (θ) = 2l + 1 (l + m)! Plm (cos(θ)). 2π)) has a purely discrete spectrum given σ(L2 ) = {l(l + 1)l ∈ N0 }. so are the polynomials. (10.9. (10. Every continuous function can be approximated by polynomials (in the sup norm and hence in the L2 norm) and since the continuous functions are dense. which can be found in any book on special functions. Angular momentum 183 contains every polynomial. . . sin(θ)dθ) ⊗ L2 ((0. if we can show that Ylm form a basis. The operator L2 on L2 ((0. (10.42) form an orthonormal basis and satisfy L2 Ylm = l(l + 1)Ylm and L3 Ylm = mYlm . m + 1.41) (10. But this follows as in the proof of Lemma 1. ϕ) = Θlm (θ)Φm (ϕ) = 2l + 1 (l + m)! P (cos(θ))eimϕ . Proof. r2 (10.3.
we have to take the boundary condition generated by u(r) = r.51) Using the unitary transformation L2 ((0. We know at least D(Al ) ⊆ D(τ ) and since D(H) = D(H0 ) it suﬃces to consider the case V = 0. ∞)). ∞). limr→0 (f (r) − rf (r)) = 0 if l = 0}. that is.49) τl = − ˜ R(r) → u(r) = rR(r). ˜ l(l + 1) 1 d 2d r + + V (r) r2 dr dr r2 D(Hl ) ⊆ L2 ((0. ∞). ∞).4. The eigenvalues of the hydrogen atom Now we want to use the considerations from the previous section to decompose the Hamiltonian of the hydrogen atom. r2 dr). case only if l > 0. r2 dr) → L2 ((0. d2 l(l + 1) + + V (r) dr2 r2 (10. we can even admit any spherically symmetric potential V (x) = V (x) with V (r) ∈ L∞ (R) + L2 ((0.8.48) where ˜ Hl R(r) = τl R(r). we need an additional boundary condition at 0 if l = 0. where I = (0.m (10. r2 dr)} reduce our operator H = H0 + V . However. It remains to investigate this operator. our operator transforms to Al f = τl f. ∞ The important observation is that the spaces Hlm = {ψ(x) = R(r)Ylm (θ. (10.p. Oneparticle Schr¨dinger operators o 10. Since we need R(r) = u(r) to r be bounded (such that (10. ∞).50) D(Al ) ⊆ L2 ((0. (10. Theorem 10.p. . r2 dr).23) is in the domain of H0 ). ∞)).46) (10. ∞). ϕ)R(r) ∈ L2 ((0. case at ∞ for any l ∈ N0 . σess (Al ) = [0. Thus we are in the l. Hence it remains to compute the domain. Hence H = H0 + V = l.47) ˜ Hl . ∞). at 0 we are in the l. The domain of the operator Al is given by D(Al ) = {f ∈ L2 (I) f. f ∈ AC(I). ∞). (10. In this case the solutions of −u (r) + l(l+1) u(r) = 0 are given by r2 u(r) = αrl+1 + βr−l . Moreover. In fact. By construction of Al we know that it is selfadjoint and satisﬁes σess (Al ) = [0. τl = − (10. τ f ∈ L2 (I).52) Proof.184 10.
drk Note that Lk is a polynomial of degree j − k. the lowest eigenvalue E0 = − γ4 is simple and the corresponding eigenfunction ψ000 (x) = γ 3 −γr/2 e 43 π 2 is positive.9. (10. (10. The eigenvalues of the hydrogen atom 185 Finally let us turn to some explicit choices for V .60) ψnlm (x) = Rnl (r)Ylm (x). (10. The eigenvalues of H (1) are explicitly given by 2 γ . r) = jl ( zr) and ub (z. The second case is that of our Coulomb potential γ γ > 0. r where we will try to compute the eigenvalues plus corresponding eigenfunctions.62) In particular.4. 2(n + 1) An orthonormal basis for the corresponding eigenspace is given by (10. .61) e γr − 2(n+1) L2l+1 ( n+l+1 γr ). The simplest case is V = 0 in this case the solutions of l(l + 1) − u (r) + u(r) = zu(r) (10. j Lk (r) = j Theorem 10. where Rnl (r) = γ 3 (n − l)! 2n3 ((n + l + 1)!)3 γr n+1 l (10. n+1 (10.57) V (r) = − .58) dr and the associated Laguerre polynomials dk Lj (r). n ∈ N0 .55) r dr r In particular.10.53) r2 are given by the spherical Bessel respectively spherical Neumann functions √ √ u(r) = α jl ( zr) + β nl ( zr). r) = jl ( zr) + inl ( zr) are the functions which are square integrable and satisfy the boundary condition (if any) near a = 0 and b = ∞. √ √ √ (10. where the corresponding diﬀerential equation can be explicitly solved. respectively.59) En = − (10.54) where 1 d l sin(r) jl (r) = (−r)l . It turns out that they can be expressed in terms of the Laguerre polynomials dj Lj (r) = er j e−r rj (10.56) ua (z.
Introducing new variables x = −λ r and v(x) = xe u( √x ) l+1 −λ this equation transforms into γ − (l + 1). Since all eigenvalues are negative. (10. 10. Otherwise we have vj ≥ (2−ε)j j! for j suﬃciently large. (10.64) xv (x) + 2(l + 1 − x)v (x) + 2nv(x) = 0. In this case u(r) is square integrable and hence an γ eigenfunction corresponding to the eigenvalue λn = −( 2(n+l+1) )2 . But ˜ ˜ j! then v (x) ≥ exp((2 − ε)x) and thus the corresponding u(r) is not square ˜ integrable near −∞.5. If n ∈ N0 . (10.186 10. The only problem is to show that we have found all eigenvalues. v0 = 1.66) (j + 1)(j + 2(l + 1)) and thus j−1 1 2(k − n) vj = . Substituting the ansatz (10.. Hence by adding a polynomial j to v(x) we can get a function v (x) such that vj ≥ (2−ε) for all j. Thus we need to ﬁnd those values of λ for which it is square integrable near +∞. Oneparticle Schr¨dinger operators o Proof. n = √ 2 −λ Now let us search for a solution which can be expanded into a convergent power series ∞ v(x) = j=0 vj xj . Since the laws of physics state that a quantum system will transfer energy to its surroundings (e. an atom emits radiation) until it eventually reaches its ground state. Nondegeneracy of the ground state The lowest eigenvalue (below the essential spectrum) of a Schr¨dinger opero ator is called ground state. this state is in some sense .65) into our diﬀerential equation and comparing powers of x gives the following recursion for the coeﬃcients 2(j − n) vj+1 = vj (10. then vj = 0 for j > n and v(x) is a polynomial.63) − u (r) + ( r2 r √ x for λ < 0.g.65) The corresponding u(r) is square integrable near 0 and satisﬁes the boundary condition (if any). It is a straightforward calculation to check that Rnl are indeed eigenγ functions of Al corresponding to the eigenvalue −( 2(n+1) )2 and for the norming constants we refer to any book on special functions. we need to look at the equation l(l + 1) γ − )u(r) = λu(r) (10.67) j! k + 2(l + 1) k=0 Now there are two cases to distinguish.
Aψ ≤ ψ. then it is simple and the corresponding eigenfunction is strictly positive.. Convolution with a strictly positive function is positivity improving. Aψ− = ( ψ. By (7.42) and (7. We call f strictly positive if f > 0 a.10.5. To set the stage let us introduce some notation. (10. Aψ− > 0. Aψ and thus 1 ψ+ . positivity improving and real (i. If A is an eigenvalue. Ag ≥ 0 (> 0) for f. So we need a positivity improving operator.e. Proof. g ≥ 0.68) that is. ψ.e. In particular. Multiplication by a positive function is positivity preserving (but not improving). Let H = L2 (Rn ). λ < 0 are since they are given by convolution . Aψ ) = 0. then it is no restriction to assume that ψ is real (since A is real both real and imaginary part of ψ are eigenfunctions as well). (10.e. This is quite surprising since the corresponding classical mechanical system is not.69) 4 Consequently ψ− = 0 or ψ+ = 0 since otherwise Aψ− > 0 and hence also ψ+ . t > 0 and Rλ (H0 ). We call f ∈ L2 (Rn ) positive if f ≥ 0 a. and f = 0. Let ψ be an eigenfunction. Then by Aψ = Aψ+ − Aψ−  ≤ Aψ+ + Aψ− = Aψ we have A = ψ. Example. Theorem 10. Without restriction ψ = ψ+ ≥ 0 and since A is positivity increasing we even have ψ = A −1 Aψ > 0. it maps real functions to real functions) operator. Clearly A is positivity preserving (improving) if and only if f. Aψ ≤ A . Nondegeneracy of the ground state 187 the most important state. We ﬁrst show that positivity improving operators have positive eigenfunctions.10. the hydrogen atom is stable in the sense that there is a lowest possible energy. Suppose A ∈ L(L2 (Rn )) is a selfadjoint. A bounded operator A is called positivity preserving if f ≥ 0 implies Af ≥ 0 and positivity improving if f ≥ 0 implies Af > 0. the electron could fall into the nucleus! Our aim in this section is to show that the ground state is simple with a corresponding positive eigenfunction. Note that it suﬃces to show that any ground state eigenfunction is positive since nondegeneracy then follows for free: two positive functions cannot be orthogonal. We have seen that the hydrogen atom has a nondegenerate (simple) ground state with a corresponding positive eigenfunction. Aψ ≤ ψ. We assume ψ = 1 and denote by ψ± = f ±2 f  the positive and negative parts of ψ. Aψ = ψ.43) both E−tH0 . Aψ − ψ..
We ﬁrst show that e−tH . (10. Oneparticle Schr¨dinger operators o with a strictly positive function. RH (λ)ψ = 0 eλt ϕ. If we set Vn = V χ{x V (x)≤n} .11. Now it remains to use (7.71) k→∞ Since we ﬁnally obtain that e−tH0 leaves N (ψ) invariant. Next I claim that for ψ positive the closed set N (ψ) = {ϕ ∈ L2 (Rn )  ϕ ≥ 0. (10. Theorem 10. If ψ is an eigenfunction of H corresponding to E0 it is an eigenfunction of RH (λ) corresponding to E01 and the claim follows since RH (λ) = −λ 1 . ψ positive.28. In other words. that is etVn ϕ ∈ N (ψ). E0 −λ .23. Proof. Suppose H = H0 + V is selfadjoint and bounded from ∞ below with Cc (Rn ) as a core. If ϕ ∈ N (ψ) we have by ≥ 0 that = 0. it is simple and the corresponding eigenfunction is strictly positive. t > 0. both etVn and e−tH e leave N (ψ) invariant and invoking again Trotter’s formula the same is true for k t t e−t(H−Vn ) = slim e− k H e k Vn . Our hope is that this property carries over to H = H0 + V . e−tH ψ dt > 0.70) is just {0}. is positivity preserving. then Vn is bounded and Hn = H0 + Vn is positivity preserving by the Trotter product formula since both e−tH0 and e−tV are. which shows that e−tH is at least positivity preserving (since 0 cannot be an eigenvalue of e−tH it cannot map a positive function to 0). λ < E0 . but this operator is positivity increasing and thus N (ψ) = {0}. ∞ Moreover. loc s Hence e−tHn → e−tH by Theorem 6. we have Hn ψ → Hψ for ψ ∈ Cc (Rn ) (note that necessarily sr V ∈ L2 ) and hence Hn → H in strong resolvent sense by Lemma 6. Hence tVn ϕe−sH ψ = 0. e−sH ψ = 0∀s ≥ 0} e−sH ψ ϕe−sH ψ (10. If E0 = min σ(H) is an eigenvalue.41) which shows ∞ s e−t(H−Vn ) → e−tH0 ϕ.188 10. ϕ. So RH (λ) is positivity increasing for λ < E0 .72) for ϕ.
Chapter 11 Atomic Schr¨dinger o operators 11. x1.1. it does not tell us whether H is selfadjoint or not.4) 189 .1 . (11.3 ) and those corresponding to the second particle by x2 = (x2.k (xj − xk ).1 .3) √ γ/ 2 H = (−∆1 ) + (−∆2 − ). x2 ). x2. y2  (11. If we assume that the interaction is again of the Coulomb type. which will give us a feeling for how the many particle case diﬀers from the one particle case and how the diﬃculties can be overcome. D(H) = H 2 (R6 ). x1.2 does not allow singularities for n ≥ 3. the Hamiltonian is given by γ .2 . Let 1 (y1 .2 . y2 ) = √ 2 then H reads in this new coordinates I I −I I (x1 .1) We ﬁrst consider the case of two particles. (11.3 ). x2. (11.2) H = −∆1 − ∆2 − x1 − x2  Since Theorem 10. It is given by N N H=− j=1 ∆j + j<k Vj. We denote the coordinates corresponding to the ﬁrst particle by x1 = (x1. Selfadjointness In this section we want to have a look at the Hamiltonian corresponding to more than one interacting particle.
It is not hard to translate this intuitive ideas into a rigorous proof. where λ0 = −γ 2 /8 is the smallest eigenvalue of the second operator if the forces are attracting (γ ≥ 0) and λ0 = 0 if they are repelling (γ ≤ 0). In the last coordinates this corresponds to a bound state of the second operator. Secondly. the ﬁrst part corresponds to the center of mass motion and the second part to the relative motion. In particular. Since both can be arbitrarily small (but positive). Moreover. we have −∆1 ≥ 0 respectively −∆2 − ∞) γ/( 2y2 ) ≥ λ0 and hence H ≥ λ0 . ∞) ⊆ σess (H). γ≤0 (11.√ ⊆ σess (H). 0. To avoid clumsy notation. Hence. one ﬁrst has to remove the center of mass motion. H is selfadjoint and semibounded √ for any γ ∈ R. ∞). this means that the movement of the whole system is somehow unbounded. There are two possibilities for this. it is the sum of a free particle plus a particle in an external Coulomb ﬁeld.5) Clearly. we will restrict ourselves to the case of one atom with N electrons whose nucleus is ﬁxed at the origin. Atomic Schr¨dinger operators o In particular. you might suspect that γ/( 2y2 ) is relatively compact with respect to −∆1 − ∆2 in L2 (R6 ) since it is with respect to −∆2 √ in L2 (R6 ). From a physics point of view. Conversely. both particles are far away from each other (such that we can neglect the interaction) and the energy corresponds to the sum of the kinetic energies of both particles. γ ≥ 0 . Let ψ1 (y1 ) be a Weyl sequence corresponding to λ ∈ [0. both particles remain close to each other and move together. Hence we expect [λ0 . Let us look at this problem from the physical view point. If λ ∈ σess (H). ψ1 (y1 )ψ2 (y2 ) is a Weyl sequence corresponding to λ + λ0 for H and thus [λ0 . we expect [0. Thus we obtain σ(H) = σess (H) = [λ0 . this is not true! This is due to the fact that γ/( 2y2 ) does not vanish as y → ∞. In particular. this implies that we do not have to deal with the center of mass motion . Firstly. in order to reveal the physics. the physically relevant information is the spectrum of the operator √ −∆2 − γ/( 2y2 ) which is hidden by the spectrum of −∆1 . √ Using that γ/( 2y2 ) has (−∆2 )bound 0 in L2 (R3 ) it is not hard to see that the same is true for the (−∆1 − ∆2 )bound in L2 (R6 ) (details will follow in the next section). Then. ∞) for −∆1 and √ ψ2 (y2 ) be a Weyl sequence corresponding to λ0 for −∆2 −γ/( 2y2 ). However. ∞) ⊆ σess (H).190 11. λ0 = −γ 2 /8.
j = ne. . The HVZ theorem 191 encountered in our example above. Hence we obtain d 2 Rn Vk ψ 2 ≤ a  j=1 n ˆ p2 ψ(p)2 dn p + b2 ψ j ˆ p2 ψ(p)2 dn p + b2 ψ j j=1 2 2 ≤ a2 Rn  2 = a2 H0 ψ + b2 ψ 2 . H = H0 + is selfadjoint and k ∞ (Rn ) is C0 Vk (y (k) ). Let Vk ∈ L∞ (Rd ) + L2 (Rd ). d ≤ 3. D(H) = H 2 (Rn ). (11.9) d 2 2 j=1 ∂ /∂ xj .11. where Vne describes the interaction of one electron with the nucleus and Vee describes the interaction of two electrons.2. Now let ψ ∈ S(Rn ). However. there is still something we . This will follow from Kato’s theorem. γj > 0. . xd ) since such transformations leave both the scalar product of L2 (Rn ) and H0 invariant. Explicitly we have γj Vj (x) = . (11. obtained by letting y (k) be the ﬁrst d coordinates of a unitary transform of Rn . then Vk ψ where ∆1 = 2 ≤ a2 Rn ∆1 ψ(x)2 dn x + b2 Rn ψ(x)2 dn x. (11. In particular. . ee.8) Proof. Theorem 11. The HVZ theorem The considerations of the beginning of this section show that it is not so easy to determine the essential spectrum of H (N ) since the potential does not decay in all directions as x → ∞.10) which implies that Vk is relatively bounded with bound 0. . 11. Then Vk is H0 bounded with H0 bound 0.7) x We ﬁrst need to establish selfadjointness of H (N ) .2.6) D(H (N ) ) = H 2 (R3N ). It suﬃces to consider one k. (11.1 (Kato). The Hamiltonian is given by N N N N H (N ) = − j=1 ∆j − j=1 Vne (xj ) + j=1 j<k Vee (xj − xk ). After a unitary transform of Rn we can assume y (1) = (x1 . (11. n = N d. be realvalued ∞ and let Vk (y (k) ) be the multiplication operator in L2 (Rn ). by our previous lemma. a core.
∞) we will localize H (N ) on sets where either one electron is far away from the others or all electrons are far away from the nucleus. Our goal for the rest of this section is to prove this result which is due to Zhislin. (11. van Winter and Hunziker and known as HVZ theorem.j )ψr . Atomic Schr¨dinger operators o can do. If the single electron is far away from the remaining system such that there is little interaction. . let us split the system into H (N −1) plus a single electron. then (H (N ) − λ − λN −1 )ψr ≤ (H (N −1) − λN −1 )ψ N −1 + ψ N −1 + (VN − j=1 1 (−∆N − λ)ψr N −1 1 ψr (11.j )ψ N −1 ∈ j=1 1 L2 (R3N ) and ψr  → 0 pointwise as r → ∞ (by Lemma 10. (−∆N − λ)ψ 1 ≤ ε for some λ ≥ 0. the third .6). Let H (N ) be the selfadjoint operator given in (11. The proof is similar but involves some additional notation. . ∞) ⊆ σess (H (N ) ) we choose Weyl sequences for H (N −1) and −∆N and proceed according to our intuitive picture from above. ψr (xN ) = ψ 1 (xN − r). the ionization energy (i. We begin with the ﬁrst inclusion. Hence arguing as in the two electron example of the previous section we expect Theorem 11. Since the error turns out relatively compact. Since (VN − N −1 VN. .192 11. For all cases where one electron is far away it is λN −1 and for the case where all electrons are far away from the nucleus it is 0 (since the electrons repel each other). Denote the inﬁmum of the spectrum of H (N ) by λN . . In fact there is a version which holds for general N body systems.1). Then H (N ) is bounded from below and σess (H (N ) ) = [λN −1 . To prove σess (H (N ) ) ⊆ [λN −1 . Then. . Now consider 1 1 ψr (x1 . . xN ) = ψ N −1 (x1 . . . xN −1 ) ∈ H 2 (R3(N −1) ) such that ψ N −1 = 1. The idea of proof is the following.2 (HVZ). where λN −1 = min σ(H (N −1) ) < 0.e.. . In particular. (H (N −1) − λN −1 )ψ N −1 ≤ ε and ψ 1 ∈ H 2 (R3 ) such that ψ 1 = 1. To prove [λN −1 . xN −1 )ψr (xN ).11) VN. Let ψ N −1 (x1 . . the energy needed to remove one electron from the atom in its ground state) of an atom with N electrons is given by λN − λN −1 .12) where VN = Vne (xN ) and VN. the energy should be the sum of the kinetic energy of the single electron and the energy of the remaining system. .j = Vee (xN −xj ). ∞). . it remains to consider the inﬁmum of the spectra of these operators.
17) We claim that ∞ N n Uj = S 3N −1 . The second inclusion is more involved. 1 ≤ j ≤ N .4. 1]. is such that N (11. n=1 j=0 (11. suppose there is an x ∈ S 3N −1 which is not an element of this union. (11. (11. Next.3 (IMS localization formula). Proof.16) = j. [0. 1]). (11. Suppose φj ∈ C ∞ (Rn ). (11. and xj  ≥ Cx}. ψ ∈ H 2 (Rn ). since n for all n implies 0 = x − x  = x  for some j > 1.14) then N ∆ψ = j=0 φj ∆φj ψ − ∂φj 2 ψ. j=0 x ∈ Rn . say j = 1. We begin with a localization formula. = j.11. and ∂φj (x) → 0 as x → ∞. which can be veriﬁed by a straightforward computation Lemma 11.2.15) Abbreviate B = {x ∈ R3N x ≥ 1}. Similarly. ∞) ⊆ σess (H (N ) ). Lemma 11. Consider the sets n Uj = {x ∈ S 3N −1  xj − x  > n−1 for all n U0 = {x ∈ S 3N −1  x  > n−1 for all }. and xj  > n−1 }.14) holds. we will choose φ0 in such a way that x ∈ supp(φ0 ) ∩ B implies that all particle are far away from the nucleus. 0 ≤ j ≤ N . say j = 2. Now we will choose φj . (H (N ) − λ − λN −1 )ψr ≤ 3ε proving [λN −1 . x ∈ U1 j 1 j . There exists functions φj ∈ C ∞ (Rn . supp(φj ) ∩ B ⊆ {x ∈ B xj − x  ≥ Cx for all supp(φ0 ) ∩ B ⊆ {x ∈ B x  ≥ Cx for all } for some C ∈ [0. in such a way that x ∈ supp(φj ) ∩ B implies that the jth particle is far away from all the others and from the nucleus. n Then x ∈ U0 for all n implies 0 = xj  for some j.13) φj (x)2 = 1. is such that (11. 0 ≤ j ≤ N . In summary. The HVZ theorem 193 term can be made smaller than ε by choosing r large (dominated convergence).18) Indeed.
j) = − =1 ∆ − =j N V + N Vk. H (N. Let V be H0 bounded with H0 bound 0 and suppose that χ{xx≥R} V RH0 (z) → 0 as R → ∞.j) = Vj + =j Vj. Let ψn converge to 0 weakly. x ∈ S 3N −1 . Note that ψn ≤ M for some M > 0. Extend φj (x) to a smooth function from 3N \{0} to [0. [0.20) ˜ and pick a function φ ∈ C ∞ (R3N . k. Then ˜ ˜ ˜ φ + (1 − φ)φj φj = (11.24) a H0 φRH0 (z)ψn + b φRH0 (z)ψn . By our localization formula we have N N H (N ) = j=0 φj H (N. =j V (N.21) N ˜ ˜ ˜ φ + (1 − φ)φ =0 are the desired functions. j (11.0) = =1 V (11. 1] by R ˜ ˜ φj (λx) = φj (x). (11. V (N. [0.194 11.19) for n suﬃciently large. . λ > 0. Atomic Schr¨dinger operators o Proceeding like this we end up with x = 0. k< .5. Proof. The gradient tends to zero since φj (λx) = φj (x) for λ ≥ 1 and x ≥ 1 which implies (∂φj )(λx) = λ−1 (∂φj )(x). Choose ∞ φ ∈ C0 (Rn .j) + ∂φj 2 .0) = − =1 ∆ + k< Vk. It is wellknown that there is a partition of unity ˜ ˜ φj (x) subordinate to this cover.j) φj + K.23) To show that our choice of the functions φj implies that K is relatively compact with respect to H we need the following Lemma 11. a contradiction. Then V RH0 (z)ψn ≤ ≤ (1 − φ)V RH0 (z)ψn + V φRH0 (z)ψn (1 − φ)V RH0 (z) ∞ ψn + (11. 1]) with support inside the unit ball which is 1 in a neighborhood of the origin. .22) where N N N N N H (N. Then V is relatively compact with respect to H0 . K= j=0 φ2 V (N. . It suﬃces to show that V RH0 (z)ψn converges to 0. 1]) such that it is one for x ≤ R. By compactness of S 3N −1 we even have N n Uj = S 3N −1 j=0 (11.
are all of the form H (N −1) plus one particle which does not interact with the others and the nucleus.11. (11. ∞) which ﬁnishes the proof of the HVZ theorem. we can also treat molecules if we assume that the nuclei are ﬁxed in space. The terms φj V (N. The HVZ theorem 195 By assumption.28) In particular. (11.k ≥ 0 and hence N ψ.j) − λN −1 ≥ 0. 1 ≤ j ≤ N . Since the operators H (N. (H (N ) − K − λN −1 )ψ = j=0 φj ψ. the same is true for the second term choosing a small. (11.22. hence they are H0 compact by Lemma 7.10. we can take products of one particle eigenfunctions to show that (11. (11. The terms ∂φj 2 are bounded and vanish at ∞. if γee = 0 (no electron interaction). the last term can also be made smaller than ε by choosing n large since φ is H0 compact.j) are relatively compact by the lemma and hence K is relatively compact with respect to H0 . n.25) Thus we obtain the remaining inclusion σess (H (N ) ) = σess (H (N ) − K) ⊆ σ(H (N ) − K) ⊆ [λN −1 .29) 2 Note that there can be no positive eigenvalues by the virial theorem. m ∈ N. This even holds for arbitrary N .j) − λN −1 )φj ψ ≥ 0. (11.27) 4 Moreover.j) . By the HVZ theorem and the considerations of the previous section we have 2 γne . (H (N. the ﬁrst term can be made smaller than ε by choosing R large. Moreover. we see 2 γne . Finally. Note that the same proof works if we add additional nuclei at ﬁxed locations. since the electron interaction term is positive. That is. let us consider the example of Helium like atoms (N = 2). K is also relatively compact with respect to H (N ) since V (N ) is relatively bounded with respect to H0 . there are eigenvalues embedded in the essential spectrum in this case. In particular H (N ) − K is selfadjoint on H 2 (R3N ) and σess (H (N ) ) = σess (H (N ) − K).2. we have H (0) ≥ 0 since Vj. Moreover. Finally.26) σess (H (2) ) = [− 2 − γne 1 1 + 4n2 4m2 ∈ σp (H (2) (γee = 0)). Next. ∞).30) . 1 ≤ j ≤ N . By Lemma 6. H (2) ≥ − σp (H (N ) ) ⊂ (−∞. 0). we have H (N.
.
Chapter 12 Scattering theory 12. Abstract theory In physical measurements one often has the following situation.1.2) and motivated by this we deﬁne the wave operators by D(Ω± ) = {ψ ∈ H∃ limt→±∞ eitH e−itH0 ψ} .1) X ψ+ (t)$$$$ Rewriting this condition we see 0 = lim t→±∞ e−itH ψ(0) − e−itH0 ψ± (0) = lim t→±∞ ψ(0) − eitH e−itH0 ψ± (0) (12. A particle is shot into a region where it interacts with some forces and then leaves the region again. (12. Ω± ψ = limt→±∞ eitH e−itH0 ψ (12. Outside this region the forces are negligible and hence the time evolution should be asymptotically free. Hence one expects asymptotic states ψ± (t) = exp(−itH0 )ψ± (0) to exist such that ψ(t) − ψ± (t) → 0 as $$ ¡ $$$$ $ ¡ $$$ ¡ $$$ ¡ ¡ ψ(t) ¡ ¡ ¡ ¡ ψ− (t) ¡ ¡ ¡ ¡ ! ¡ ¡ t → ±∞.3) 197 .
D(Ω± ) reduces exp(−itH0 ) and Ran(Ω± ) reduces exp(−itH). diﬀerentiating (12. Next. (12. The sets D(Ω± ) and Ran(Ω± ) are closed and Ω± : D(Ω± ) → Ran(Ω± ) is unitary. ϕ ∈ D(Ω± ). .8) It is interesting to know the correspondence between incoming and outgoing states. ψ = 0.2. Moreover.198 12. + D(S) = {ψ ∈ D(Ω− )Ω− ψ ∈ Ran(Ω+ )}. ψ ∈ D(Ω± ). D(Ω± ) is invariant under exp(−itH0 ) and Ran(Ω± ) is invariant under exp(−itH).5) and hence Ω± e−itH0 ψ = e−itH Ω± ψ. Moreover. (12. Moreover. if ψ ∈ D(Ω± )⊥ then ϕ. it is called a scattering state. If a state ψ has both. In summary. We ﬁrst show a criterion due to Cook. that is. Theorem 12.1 the intertwining property of the wave operators. D(H0 ) ∩ D(S). By construction we have Ω± ψ = lim t→±∞ eitH e−itH0 ψ = lim t→±∞ ψ = ψ (12. Lemma 12. exp(−itH0 )ψ = exp(itH0 )ϕ. ψ ∈ D(Ω± ) ∩ D(H0 ). Consequently.9) Note that we have D(S) = D(Ω− ) if and only if Ran(Ω− ) ⊆ Ran(Ω+ ) and Ran(S) = D(Ω+ ) if and only if Ran(Ω+ ) ⊆ Ran(Ω− ).6) In addition. S is unitary from D(S) onto Ran(S) and we have H0 Sψ = SH0 ψ. (12. note that this whole theory is meaningless until we can show that D(Ω± ) are nontrivial.7) Hence D(Ω± )⊥ is invariant under exp(−itH0 ) and Ran(Ω± )⊥ is invariant under exp(−itH). observe that t→±∞ lim eitH e−itH0 (e−isH0 ψ) = lim e−isH (ei(t+s)H e−i(t+s)H0 ψ) t→±∞ (12. ψ ∈ Ran(Ω+ ) ∩ Ran(Ω− ). (12.4) and it is not hard to see that D(Ω± ) is closed. The subspaces D(Ω± ) respectively Ran(Ω± ) reduce H0 respectively H and the operators restricted to these subspaces are unitarily equivalent Ω± H0 ψ = HΩ± ψ. (12. Hence we deﬁne the scattering operator S = Ω−1 Ω− .6) with respect to t we obtain from Theorem 5. Scattering theory The set D(Ω± ) is the set of all incoming/outgoing asymptotic states ψ± and Ran(Ω± ) is the set of all states which have an incoming/outgoing asymptotic state. interchanging the roles of H0 and H amounts to replacing Ω± by Ω−1 and hence Ran(Ω± ) is ± also closed.1.10) However. Moreover.
Since we want to use Cook’s lemma. that is. Proof. Abstract theory 199 Lemma 12. Since such functions are dense.17) and that they are complete if.12) in this case. (12. If ∞ (H − H0 ) exp( itH0 )ψ dt < ∞.16) 4π 1 and thus any such ψ is in D(Ω+ ). As a simple consequence we obtain the following result for Schr¨dinger o operators in R3 Theorem 12. s > 0. that is. Proof. Moreover. Hence for ψ ∈ Hpp (H0 ) it is easy to check whether it is in D(Ω± ) or not and only the continuous subspace is of interest. We will say that the wave operators exist if all elements of Hac (H0 ) are asymptotic states. Similarly for Ω− . By the intertwining property ψ is an eigenfunction of H0 if and only if it is an eigenfunction of H. (12.14) for given ψ ∈ D(H0 ).15) at least for ψ ∈ L1 (R3 ). x)2 dx. (12. we need to estimate V ψ(s) 2 = R3 V (x)ψ(s. Hac (H) ⊆ Ran(Ω± ). this implies ∞ 1 V ψ(s) ds ≤ 3/2 ψ 1 V (12. Invoking (7.13) which holds for ψ ∈ D(H0 ).11) then ψ ∈ D(Ω± ). Moreover.1.4.3 (Cook). all elements of Hac (H) are scattering states. Suppose D(H) ⊆ D(H0 ). we obtain D(Ω+ ) = H. 0 ∞ ψ ∈ D(H0 ).18) . ψ(s) = exp(isH0 )ψ. Hac (H0 ) ⊆ D(Ω± ) (12. Suppose H0 is the free Schr¨dinger operator and H = o H0 + V with V ∈ L2 (R3 ). we even have (Ω± − I)ψ ≤ 0 (H − H0 ) exp( itH0 )ψ dt (12. then the wave operators exist and D(Ω± ) = H. in addition. The result follows from eitH e−itH0 ψ = ψ + i 0 t exp(isH)(H − H0 ) exp(−isH0 )ψds (12.12.34) we get 1 ψ V ψ(s) ≤ ψ(s) ∞ V ≤ (4πs)3/2 1 V . (12. respectively.
But how should incoming respectively outgoing be deﬁned for ψ ∈ H? Well incoming (outgoing) means that the expectation of x2 should decrease (increase).20) where D is the dilation operator introduced in (10. This physical picture suggests that we should be able to decompose ψ ∈ H into an incoming and an outgoing part. we will show below that the tail decays faster then any inverse power of t. d Eψ (x(t)2 ) = ψ(t). since we have assumed the mass to be two). Since the dilation group acts on x only. Dψ(t) . x2 ]ψ(t) = 4 ψ(t). Set x(t)2 = exp(iH0 t)x2 exp(−iH0 t).22) and hence Ff (D) = f (−D)F. 12. Note FD = −DF (12.200 12. (12.23) . Hence it is natural to consider ψ ∈ Ran(P± ). (12.21) as outgoing respectively incoming states. In particular asymptotic completeness implies Hsc (H) = ∅ since H restricted to Ran(Ω± ) is unitarily equivalent to H0 . b2 ). ±∞)). o In this later case the wave operators exist if D(Ω± ) = H. then. To say more we will look for a transformation which maps D to a multiplication operator. it seems reasonable to switch to polar coordinates x = rω. If we project a state in Ran(P± ) to energies in the interval (a2 .2. ω) ∈ R+ × S n−1 . In fact. we expect that it cannot be found in a ball of radius proportional to at as t → ±∞ (a is the minimal velocity of the particle. In these coordinates we have U (s)ψ(eρ ω) = e−ns/2 ψ(e(ρ−s) ω) (12. Scattering theory If we even have Hc (H) ⊆ Ran(Ω± ). Incoming and outgoing states In the remaining sections we want to apply this theory to Schr¨dinger opo erators. Our ﬁrst goal is to give a precise meaning to some terms in the intuitive picture of scattering theory introduced in the previous section. i[H0 . Since U (s) essentially transforms r into r exp(s) we will replace r by ρ = ln(r). We ﬁrst collect some properties of D which will be needed later on. dt ψ ∈ S(Rn ). and asymptotically complete if Hc (H) = Ran(Ω± ). (12. abbreviating ψ(t) = e−itH0 ψ. P± = PD ((0. (t.19) they are called asymptotically complete. We will be mainly interested in the case where H0 is the free Schr¨dinger operator and hence Hac (H0 ) = H.11). they are complete if Hac (H) = Ran(Ω± ).
∞ Lemma 12. −R))Kt. Incoming and outgoing states 201 and hence U (s) corresponds to a shift of ρ (the constant in front is absorbed by the volume element).26) (12. ∞))ψ ˆ = Kt. (1 + t)N ±t ≥ 0. that is. ±∞)) ≤ respectively.28) and that S(Rn ) is a core for D.30) where Kt.5.32) .x . Thus D corresponds to diﬀerentiation with respect to this coordinate and all we have to do to make it a multiplication operator is to take the Fourier transform with respect to ρ.25) S n−1 . (12.2. Introducing ψ(t. N ∈ N there is a constant C such that χ{x x<2at} e−itH0 f (H0 )PD ((±R. In particular we have P+ + P− = I. PD ((−∞. r−iλ ψ(rω)r n −1 2 (12. (12. FPD ((R. For any R ∈ R. (12. This leads us to the Mellin transform M : L2 (Rn ) → L2 (R × S n−1 ) ψ(rω) 1 → (Mψ)(λ. t > 0. C . (12. (Mψ)(λ. the remaining one being similar.24) dr By construction. Suppose f ∈ Cc (R) with supp(f ) ⊂ (a2 .12. ω) = √ 2π 0 ∞ . M is unitary. (12. σsc (D) = σpp (D) = ∅ (12. Moreover.31) ≤ const . ∞))ψ(x) = Kt. ω)2 dλdn−1 ω = R S n−1 R+ S n−1 ψ(rω)2 rn−1 drdn−1 ω.x 2 1 2 ei(tp −px) f (p2 )∗ . From this it is straightforward to show that σ(D) = σac (D) = R.29) Proof. (1 + t)2N (12.x . −R))ψ . We prove only the + case. Consider ψ ∈ S(Rn ).27) −isλ where dn−1 ω is the normalized surface measure on M −1 U (s)M = e and hence M−1 DM = λ. Using the Mellin transform we can now prove Perry’s estimate [11].x (p) = we see that it suﬃces to show PD ((−∞. x) = e−itH0 f (H0 )PD ((R. b2 ) for some a. b > 0. n/2 (2π) for x < 2at.
r ∈ (a.6. Estimating the derivative of α we see α (r) = 2tr + ωx − λ/r > 0. ψ(t) will eventually leave every compact ball (at least on the average). (12. we expect such potentials to be asymptotically complete.34) ∞ ˜ with f (r) = f (r2 )∗ rn/2−1 ∈ Cc ((a2 .x )(λ. ∞ Corollary 12. (12. Hence we expect that the time evolution will asymptotically look like the free one for ψ ∈ Hc if the potential decays suﬃciently fast. t as above. .3.37) (the last step uses integration by parts) for λ. t as above. Scattering theory Now we invoke the Mellin transform to estimate this norm R PD ((−∞.x )(λ. ∞)) and R ∈ R. r ∈ (a.x Since 2 = −∞ S n−1 (MKt. Proof. −R))Kt.33) (MKt. Abbreviating PD = PD ((±R.202 12. Taking t → ∞ the ﬁrst term goes to zero by our lemma and the second goes to zero since χψ → ψ. (12. b). ±∞)) and χ = χ{x x<2at} we have PD f (H0 )e−itH0 ψ ≤ χeitH0 f (H0 )∗ PD ψ + f (H0 ) (I−χ)ψ . b2 )). Hence we can ﬁnd a constant such that of f const 1 ≤ .x )(λ. Then the operator PD ((±R. b). 12. This ﬁnishes the proof.39) since A = A∗ .38) (MKt.34) ∞ 0 1 d iα(r) const ˜ e dr ≤ f (r) α (r) dr 1 + λ + t ∞ 0 ˜ f (r)eiα(r) dr .35) for λ ≤ −R and t > where ε is the distance of a to the support ˜. by iterating the last estimate we see const (12. ±∞))f (H0 ) exp(−itH0 ) converges strongly to 0 as t → ∞. ω) ≤ (1 + λ + t)N for any N ∈ N and t ≥ 0 and λ ≤ −R. R(2εa)−1 . By increasing the constant we can even assume that it holds for t ≥ 0 and λ ≤ −R. Moreover. ω)2 dλdn−1 ω.36) α (r) 1 + λ + t and λ. (12. α(r) = tr2 + rωx − λ ln(r). Suppose that f ∈ Cc ((0. Schr¨dinger operators with short range o potentials By the RAGE theorem we know that for ψ ∈ Hc . ω) = ˜ f (r)eiα(r) dr (12. In other words. 1 (2π)(n+1)/2 0 ∞ (12. Using this we can estimate the integral in (12.
12. Schr¨dinger operators with short range potentials o 203 Suppose V is relatively bounded with bound less than one. Proof. 1]) such that φ(x) = 0 for 0 ≤ x ≤ 1/2 and φ(x) = 0 for 1 ≤ x. [0. Using ˜ h2 (r) = φr V RH0 (z) . and φr (x) = φ(x/r).46) Lemma 12.45) (12.7. r ˜ ˜ Hence h2 is integrable if h1 is. φr ]RH0 (z) = RH0 (z)(∆φr + (∂φr )∂)RH0 (z) and ∆φr = φr/2 ∆φr . Then it is not hard to see that hj is integrable if and ˜ only if hj is integrable. (12.40) ˜ h1 (r) = V RH0 (z)φr . then RH (z) − RH0 (z) is compact. by our short range condition it converges in norm to RH (z)V RH0 (z) = RH (z) − RH0 (z) as r → ∞ (at least for some subsequence). If V is short range. (12. Pick φ ∈ Cc (Rn .42) [RH0 (z). Moreover.47) . Invoking the ﬁrst resolvent formula φr V RH0 (z) ≤ φr V RH0 (z0 ) I − (z − z0 )RH0 (z) ﬁnishes the proof. Moreover.3. The operator RH (z)V (I−χr )RH0 (z) is compact since (I−χr )RH0 (z) is by Lemma 7. φr ] = −RH0 (z)[H0 (z).8. c˜ c˜ 2c ˜ ˜ ˜ ˜ h1 (r) ≤ h2 (r) + h1 (r/2) ≤ h2 (r) + h2 (r/2) + 2 h1 (r/4) r r r ˜ ˜ shows that h2 is integrable if h1 is.10 and RH (z)V is bounded by Lemma 6. r ≥ 1. r ≥ 1. (12. The function h1 is integrable if and only if h2 is. ∞ Proof. Introduce h1 (r) = V RH0 (z)χr . r ≥ 0. We ﬁrst note that it suﬃces to check this for h1 or h2 and for one z ∈ ρ(H0 ).41) The potential V will be called short range if these quantities are integrable.22. where h2 (r) = χr V RH0 (z) .44) c˜ ˜ ˜ h1 (r) − h2 (r) ≤ h1 (r/2). where χr = χ{x x≥r} . (12. As a ﬁrst consequence note (12. Conversely.43) respectively (∂φr ) = φr/2 (∂φr ). (12. hj integrable for one z0 ∈ ρ(H0 ) implies hj integrable for all z ∈ ρ(H0 ). Lemma 12. ∆φr ∞ ≤ ∆φ ∂φr ∞ ≤ ∂φ ∞ /r2 we see ∞ /r 2 (12.
53) .48) Proof.5) for the ﬁrst term if we choose ψ = f (H0 )PD ((±R. we see that the wave operators exist. ∞) = σac (H0 ) ⊆ σac (H). ∞). Then lim (Ω± − I)f (H0 )P± ψn = 0.49) is integrable for a dense set of vectors ψ. (12. ∞)) and suppose ψn converges weakly to 0. By Cook’s criterion (Lemma 12. We begin by showing that the wave operators exist. D(Ω± ) = H.13) we see ∞ RH (z)(Ω± − I)f (H0 )P± ψn ≤ 0 RH (z)V exp(−isH0 )f (H0 )P± ψn dt. (12.10. And by σac (H) ⊆ σess (H) = [0. ∞ Lemma 12.204 12. we obtain ± [0. Lemma 12. (12.9. Since vectors of this form are dense. (12. Moreover.3) we need to show that V exp( itH0 )ψ ≤ V RH0 (−1) (I − χ2at ) exp( itH0 )(H0 + I)ψ + V RH0 (−1)χ2at (H0 + I)ψ (12. this is clearly not enough to prove asymptotic completeness and we need a more careful analysis.8 since χr converges strongly to 0. The second term is integrable by our short range assumption. However. To prove asymptotic completeness of the wave operators we will need that (Ω± − I)f (H0 )P± are compact. Scattering theory In particular. by Weyl’s theorem we have σess (H) = [0.52) Since RH (z)V RH0 is compact we see that the integrand RH (z)V exp(−isH0 )f (H0 )P± ψn = RH (z)V RH0 exp(−isH0 )(H0 + 1)f (H0 )P± ψn (12.51) n→∞ that is. The same is true by Perry’s estimate (Lemma 12.20 and the second part follows from part (ii) of Lemma 6. ∞). By (12. ±∞))ϕ. (Ω± − I)f (H0 )P± is compact. ∞) we even have σac (H) = [0. then so is f (H)−f (H0 ) for any f ∈ C∞ (R) and r→∞ lim (f (H) − f (H0 ))χr = 0. The main ideas are due to Enß [4].50) Since H restricted to Ran(Ω∗ ) is unitarily equivalent to H0 . Suppose RH (z)−RH0 (z) is compact. Proof. Let f ∈ Cc ((0. V short range implies that H and H0 look alike far outside. The ﬁrst part is Lemma 6.
Choosing f such that f (λ) = 1 for λ ∈ [a. (12. Introduce ϕ± (t) = f (H0 )P± ψ(t). ∞). (12. Then σac (H) = σess (H) = [0. it follows that P f (H) is also compact. pp sc Observe that replacing PH by PH the same argument shows that all nonzero eigenvalues are ﬁnite dimensional and cannot accumulate in (0. Schr¨dinger operators with short range potentials o 205 converges pointwise to 0. b) = ∅. sc Abbreviate P = PH PH ((a. b] we see that P = P f (H) is compact and hence ﬁnite dimensional. but avoids the use of Ces`ro means in our main a argument. Abbreviate ψ(t) = exp(−itH)ψ. 0 < a < b. Thus RH (z)(Ω± − I)f (H0 )P± is compact by the dominated convergence theorem. that is. (12. In particular σsc (H) ∩ (a. Since f (H) − f (H0 ) is compact. the singular part must live on Ran(Ω± )⊥ .11. b) is a ﬁnite set. Since 0 < a < b are arbitrary we even have σsc (H) ∩ (0. where f (λ) = (λ + 1)f (λ). Moreover. ∞) = ∅ and by σsc (H) ⊆ σess (H) = [0. We ﬁrst show that the singular continuous spectrum is absent.3.49) the integrand is bounded by an L1 function depending only on ψn .57) .56) ∞ for some f ∈ Cc ((0. Thus P f (H0 ) = P (I − Ω+ )f (H0 )P+ + P (I − Ω− )f (H0 )P− is compact. ∞). using the intertwining property we see that ˜ (Ω± − I)f (H0 )P± = RH (z)(Ω± − I)f (H0 )P± − (RH (z) − RH0 (z))f (H0 )P± ˜ is compact by Lemma 6. Suppose V is short range. Since H restricted to Ran(Ω± ) is unitarily equivalent to H0 (which has purely absolutely continusc ous spectrum). Now we have gathered enough information to tackle the problem of asymptotic completeness. PH Ω± = 0. showing σsc (H) ∩ (a. This is not really necessary.54) In summary we have shown Theorem 12. But a continuous measure cannot be supported on a ﬁnite set. ∞).20. ∞). By the RAGE theorem the sequence ψ(t) converges weakly to zero as t → ±∞. Choose ψ ∈ Hc (H) = Hac (H) such that ψ = f (H)ψ (12. ∞) we obtain σsc (H) = ∅.12.55) All nonzero eigenvalues have ﬁnite multiplicity and cannot accumulate in (0. arguing as in (12. σsc (H) = ∅. Now we come to the anticipated asymptotic completeness result of Enß. b)). Furthermore.
ψ(t) . Ω ϕ (t) (12. we even have t→±∞ (12. . Moreover.60) lim (Ω± − I)ϕ± (t) = 0 by Lemma 12. For further results and references see for example [3]. Ran(Ω± )⊥ is invariant under H and hence ψ(t) ∈ Ran(Ω± )⊥ implying ψ 2 = = t→±∞ t→±∞ lim ψ(t).59) (12. then ψ 2 = = = t→±∞ lim ψ(t). Ω+ ϕ+ (t) + Ω− ϕ− (t) . (12. (12. ψ(t) lim ψ(t).63) by Corollary 12. Suppose V is short range.61) t→±∞ t→±∞ By Theorem 12. ψ(t) = 0 t→±∞ (12.206 12.2.6. Now suppose ψ ∈ Ran(Ω± )⊥ . Scattering theory which satisfy t→±∞ lim ψ(t) − ϕ+ (t) − ϕ− (t) = 0.12 (Enß). then the wave operators are asymptotically complete.20.10. Invoking the intertwining property we see ψ 2 = lim P f (H0 )∗ e−itH0 Ω∗ ψ. ϕ+ (t) + ϕ− (t) lim ψ(t). Hence Ran(Ω± ) = Hac (H) = Hc (H) and we thus have shown Theorem 12.58) Indeed this follows from ψ(t) = ϕ+ (t) + ϕ− (t) + (f (H) − f (H0 ))ψ(t) and Lemma 6.62) lim P f (H0 )∗ Ω∗ ψ(t).
Part 3 Appendix .
.
• A is closed under complements.1. since a classical paradox by Banach and Tarski shows that one can break the unit ball in R3 into a ﬁnite number of (wild – choosing the pieces uses the Axiom of Choice and cannot be done with a jigsaw. rotate and translate them. 209 . If an algebra is closed under countable unions (and hence also countable intersections)..1).e. it is called a σalgebra. Hence any reasonable notion of size (i. and reassemble them to get two copies of the unit ball (compare Problem A. this turns out to be too much. A. Good references are [2] or [18].) pieces. is called an algebra. Note that ∅ ∈ A and that. Borel measures in a nut shell The ﬁrst step in deﬁning the Lebesgue integral is extending the notion of size from intervals to arbitrary sets. one which is translation and rotation invariant) cannot be deﬁned for all sets! A collection of subsets A of a given set X such that • X ∈ A. A is also closed under ﬁnite intersections. by de Morgan. • A is closed under ﬁnite unions.Appendix A Almost everything about Lebesgue integration In this appendix I give a brief introduction to measure theory. Unfortunately.
If we replace the σalgebra by an algebra A. Example. In the case X = Rn the Borel σalgebra will be denoted by Bn and we will abbreviate B = B1 . ˜ Proof. the intersection of any family of (σ)algebras {Aα } is again a (σ)algebra and for any collection S of subsets there is a unique smallest (σ)algebra Σ(S) containing S (namely the intersection of all (σ)algebra containing S). Any measure µ satisﬁes the following properties: (i) A ⊆ B implies µ(A) ≤ µ(B) (monotonicity). A and µ(A1 ) < ∞ (continuity from above). Sets in the Borel σalgebra are called Borel sets. In this case σadditivity clearly only needs to hold for disjoint sets An for which n An ∈ A. k (σadditivity). The second follows using An = An \An−1 ˜n = A1 \An and and σadditivity. (ii) µ(An ) → µ(A) if An (iii) µ(An ) → µ(A) if An A (continuity from below). If X is a topological space. the Borel σalgebra of X is deﬁned to be the σalgebra generated by all open (respectively all closed) sets. then µ(An ) = ∞. The sets in Σ are called measurable sets. Example.1. Now let us turn to the deﬁnition of a measure: A set X together with a σalgebra Σ is called a measure space. Let A ∈ P(M ) and set µ(A) to be the number of elements of A (respectively ∞ if A is inﬁnite). The third follows from the second using A ˜ µ(An ) = µ(A1 ) − µ(An ). n An ) and An A if Theorem A.210 A.1 is not superﬂuous. but µ( n An ) = µ(∅) = 0 which shows that the requirement µ(A1 ) < ∞ in the last claim of Theorem A. We will write An An+1 ⊆ An (note A = A if An ⊆ An+1 (note A = n An ). It is called the (σ)algebra generated by S. (Note that it is no restriction to assume Xj X. Note that if X = N and An = {j ∈ Nj ≥ n}. This is the so called counting measure. then µ is called a premeasure. It is called σﬁnite if there is a countable cover {Xj }∞ of X with µ(Xj ) < j=1 ∞ for all j. • µ( ∞ j=1 Aj ) ∞ j=1 µ(Aj ) = if Aj ∩ Ak = ∅ for all j. A measure µ is a map µ : Σ → [0. Almost everything about Lebesgue integration Moreover. . The ﬁrst claim is obvious. ∞] on a σalgebra Σ such that • µ(∅) = 0.) It is called ﬁnite if µ(X) < ∞.
Let us ﬁrst show how we should deﬁne µ for intervals: To every Borel measure on B we can assign its distribution function −µ((x. x=0 µ(x) = (A.A.4). b] µ(b) − µ(a). Show that the restriction to the Borel sets is a measure. . b) where µ(a−) = limε↓0 µ(a − ε). b] µ(A) = .4) A = (a. (A. In particular. A Borel measures is called outer regular if µ(A) = and inner regular if µ(A) = C⊆A. Example. For every right continuous nondecreasing function µ : R → R there exists a unique regular Borel measure µ which extends (A. which is given by Θ(A) = 1 if 0 ∈ A and Θ(A) = 0 if 0 ∈ A. Since the proof of this theorem is rather involved. A = [a.3) µ((0.1. Borel measures in a nut shell 211 A measure on the Borel σalgebra is called a Borel measure if µ(C) < ∞ for any compact set C. Suppose Θ(x) = 0 for x < 0 and Θ(x) = 1 for x ≥ 0.1) sup µ(C). Then we obtain the socalled Dirac measure at 0. µ(b) − µ(a−). A = [a. x>0 which is right continuous and nondecreasing. this gives a premeasure on the algebra of ﬁnite unions of intervals which can be extended to a measure: Theorem A. Extend this to an outer measure for all subsets of R. Conversely. we defer it to the next section and look at some examples ﬁrst.O inf µ(O) open (A.2) It is called regular if it is both outer and inner regular.C A⊆O. b) µ(b−) − µ(a). given a right continuous nondecreasing function µ : R → R we can set A = (a. compact (A. µ(b−) − µ(a−). Then the strategy is as follows: Start with the algebra of ﬁnite unions of disjoint intervals and deﬁne µ for those sets (as the sum over the intervals). 0]). x]).2. Two diﬀerent functions generate the same measure if and only if they diﬀer by a constant. But how can we obtain some more interesting Borel measures? We will restrict ourselves to the case of X = R for simplicity. x < 0 0. This yields a premeasure.
The set of rational numbers has Lebesgue measure zero: λ(Q) = 0. 7 ]∪ 9 9 3 3 9 [ 8 .212 A. Problem A. λ(C) = limn→∞ λ(Cn ) = 0. again remove 3 the middle third’s of the remaining sets to obtain C2 = [0. (Hint: How can you build up [0.e.) if the it holds on a support for µ or.. It can be shown that Borel measures on a separable metric space are always regular. The Cantor set is an example of a closed uncountable set of Lebesgue measure zero. Since Cn is compact. Proceeding like this we obtain a sequence of nesting sets Cn and the limit C = n Cn is the Cantor set. 1] 1 and remove the middle third to obtain C1 = [0. 1) equivalent if x − y is rational. .e. any single point has Lebesgue measure zero. so is C. Next. Using the ternary expansion it is extremely simple to describe: C is the set of all x ∈ [0. Suppose λ(x) = x. We will abbreviate the Lebesgue measure of a Borel set A by λ(A) = A. then the associated measure is the ordinary Lebesgue measure on R. In particular. A property is said to hold µalmost everywhere (a. A set A ∈ Σ is called a support for µ if µ(X\A) = 0. . Almost everything about Lebesgue integration Example. if the set where it does not hold is contained in a set of measure zero. 1). equivalently. Example. Show that V cannot be measurable with respect to any ﬁnite translation invariant measure on [0. It has some further interesting properties: it is totally disconnected (i. Construct the set V by choosing one representative from each equivalence class. and so has any countable union of points (by countable additivity). 1].1 (Vitali set). 1 ]∪[ 2 . Moreover. Example. 1 ]∪[ 2 . 1] whose ternary expansion contains no one’s. 1]. It is constructed as follows: Start with C0 = [0. 1) from V ?) . and thus its Lebesgue measure is λ(Cn ) = (2/3)n . 9 C0 C1 C2 C3 . Call two numbers x. In fact. which shows that C is uncountable (why?). 3 ]∪[ 2 . it contains no subintervals) and perfect (it has no isolated points). y ∈ [0. Cn consists of 2n intervals of length 3−n .
4) indeed gives rise to a premeasure. To show that it is an σalgebra let ˜ ˜ ˜ An be given and put An = k≤n An . Now we start by proving that (A. But this shows A ⊆ M (A) and hence M (A) = M for any A ∈ M. Then M(A) = Σ(A). Theorem A. Similarly for decreasing sequences.. Every σalgebra is a monotone class and the intersection of monotone classes is a monotone class. Similarly.2 we need to show how a premeasure can be extended to a measure. In order to show that it holds for any set in Σ(A). Then An is increasing and n An = n An ∈ A. Hence every collection of sets S generates a smallest monotone class M(S). If Bn is an increasing sequence of sets in M (A) then A ∪ Bn is an increasing sequence in M and hence n (A ∪ Bn ) ∈ M.2.3. Put M (A) = {B ∈ MA ∪ B ∈ M}. But does it contain any elements? Well if A ∈ A we have A ⊆ M (A) implying M (A) = M. Now A∪ n Bn = n (A ∪ Bn ) (A. Extending a premasure to a measure The purpose of this section is to prove Theorem A.2. M (A) is closed under decreasing sequences and hence it is a monotone class. Extending a premasure to a measure 213 A. is a monotone class).A. It is rather technical and should be skipped on ﬁrst reading.e.2. it suﬃces to show that the sets for which it holds is closed under countable increasing and decreasing sequences (i. As a prerequisite we ﬁrst establish that it suﬃces to check increasing (or decreasing) sequences of sets when checking wether a given algebra is in fact a σalgebra: A collections of sets M is called a monotone class if An A implies A ∈ M whenever An ∈ M and An A implies A ∈ M whenever An ∈ M. The typical use of this theorem is as follows: First verify some property for sets in an algebra A. In order to prove Theorem A. To show that we are closed under complements consider M = {A ∈ MX\A ∈ M}. Hence A ∪ B ∈ M if at least one of the sets is in A. So we know that M is an algebra. Hence M is a monotone class and must be equal to M since it contains A. If An is an increasing sequence then X\An is a decreasing sequence and X\ n An = n X\An ∈ M if An ∈ M . So M is closed under ﬁnite unions.5) shows that M (A) is closed under increasing sequences. . We ﬁrst show that M = M(A) is an algebra. Proof. Let A be an algebra.
6) whenever In ∈ A and I = k Ik ∈ A. If I is unbounded. To show outer regularity replace each point {x} by a small open interval (x + ε. µ as deﬁned in (A. (A. x − ε) and use that µ({x}) = limε↓ µ(x + ε) − µ(x−ε). We need to show µ( k Ik ) = k µ(Ik ) (A. b) by a compact one [an . we can assume any such union to consist of open intervals and points only.9) Since ε > 0 is arbitrary. b) and use µ((a. Almost everything about Lebesgue integration Lemma A.8) To get the converse inequality we need to work harder.7) which shows ∞ µ(Ik ) ≤ µ(I). By outer regularity we can cover each Ik by open interval Jk such that µ(Jk ) ≤ µ(Ik ) + 2εk . It remains to verify σadditivity.214 A. say the ﬁrst n.4) can be extended to ﬁnite unions of disjoint intervals by summing over all intervals. Similarly. . Suppose I is compact ﬁrst. Then ﬁnitely many of the Jk . we can as well assume each In is just one interval (just split In into its subintervals and note that the sum does not change by additivity). say I = [a. this shows σadditivity for compact intervals. By additivity µ is monotone and hence n n µ(Ik ) = µ( k=1 k=1 Ik ) ≤ µ(I) (A. ∞). to show inner regularity. (A. we can assume that I is just one interval (just treat each subinterval separately). Similarly.4) gives rise to a unique σﬁnite regular premeasure on the algebra A of ﬁnite unions of disjoint intervals.4. Proof. b)) = limn→∞ µ(bn ) − µ(an ) if an ↓ a and bn ↑ b. Since each In is a ﬁnite union of intervals. By additivity we can always add/subtract the end points of I and hence σadditivity holds for any bounded interval. cover I and we have n n ∞ µ(I) ≤ µ( k=1 Jk ) ≤ k=1 µ(Jk ) ≤ k=1 µ(Ik ) + ε. replace each open interval (a. First of all. bn ] ⊆ (a. k=1 (A. It is straightforward to verify that µ is well deﬁned (one set can be represented by diﬀerent unions of intervals) and by construction additive. To show regularity.
By the ﬁnite case µ(A ∩ Xj ) = µ(A ∩ Xj ) (just restrict µ. Extending a premasure to a measure 215 then given x > 0 we can ﬁnd an n such that Jn cover at least [0.3). Let An ∈ M be a monotone sequence and let On ⊇ An be open sets such 1 that µ(On ) ≤ µ(An ) + n . Suppose µ is a σﬁnite premeasure on some algebra A generating the Borel sets B. Proof. x]) − ε. µ to Xj ). Suppose there is another extension µ and consider the set ˜ S = {A ∈ Σ(A)µ(A) = µ(A)}. (A.2. Then outer (inner) regularity holds for all Borel sets if it holds for all sets in A. Note that if our premeasure is regular. (A.5 (Uniqueness of measures).12) and we are done.A.O inf µ(O) ≥ µ(A) open (A. Let An A. ˜ (A. Then 1 .13) and let M = {A ∈ Bµ◦ (A) = µ(A)}.6. so will be the extension: Lemma A. We ﬁrst assume that µ(X) < ∞. Similarly µ(An ) ≤ µ◦ (An ) ≤ µ(On ) ≤ µ(An ) + Now if An if An A. Since by assumption M contains some algebra generating B it suﬃces to proof that M is a monotone class. Set µ◦ (A) = A⊆O. This premeasure determines the corresponding measure µ uniquely (if there is one at all): Theorem A. Then there is at most one extension to Σ(A). We ﬁrst assume that µ(X) < ∞. Let µ be a σﬁnite premeasure on an algebra A.11) I claim S is a monotone class and hence S = Σ(A) since A ⊆ S by assumption (Theorem A. If An ∈ S we have µ(An ) = µ(An ) and taking limits ˜ (Theorem A. Next let An ˜ A and take again limits.10) Since x > a and ε > 0 are arbitrary we are done. To extend our result to the σﬁnite case let Xj X be an increasing sequence such that µ(Xj ) < ∞.14) n A just take limits and use continuity from below of µ. Proof.1 (ii)) we conclude µ(A) = µ(A). x] and hence n n µ(Ik ) ≥ k=1 k=1 µ(Jk ) − ε ≥ µ([a. . This ﬁnishes the ﬁnite case. Hence ˜ ˜ µ(A) = lim µ(A ∩ Xj ) = lim µ(A ∩ Xj ) = µ(A) ˜ ˜ j→∞ j→∞ (A.
216 A. covers A. etc. the sum j µ(Aj ) will ˜ n = n ⊆ A is compact with diverge and so will j=1 j µ(Kj ). Let Xj be a cover with µ(Xj ) < ∞. A n ∈ A (A. that is. Then the set Σ of all sets A satisfying the Carath´odory condition e µ∗ (E) = µ∗ (A ∩ E) + µ∗ (A ∩ E) ∀E ⊆ X (A.17) µ∗ (A) = inf n=1 µ(An ) A ⊆ n=1 An . Since Xj has ﬁnite measure.15) This settles outer regularity.7 (Extensions via outer measures). A2 = (A\A1 ) ∩ X2 . where µ◦ (A) = C⊆A.16) is a monotone class. This settles the ﬁnite case. it has the properties (Problem A. (A. we can assume Xj open. Given A we can split it into disjoint sets Aj such that Aj ⊆ Xj (A1 = A ∩ X1 .19) . So it remains to ensure that there is an extension at all. and satisﬁes µ(A) ≤ µ(O) ≤ j µ(Oj ) ≤ µ(A) + ε. If µ(X) < ∞ one can show as before that M = {A ∈ Bµ◦ (A) = µ(A)}. and • µ∗ ( ∞ n=1 An ) ≤ ∞ ∗ n=1 µ (An ) (subadditivity). Thus there are open (in X) sets Oj covering Aj such that µ(Oj ) ≤ µ(Aj ) + 2εj . For any premeasure µ we deﬁne ∞ ∞ (A. ∞] is an outer measure. Almost everything about Lebesgue integration Now let µ be arbitrary. Then the function µ∗ : P(X) → [0.2) • µ∗ (∅) = 0.).18) where the inﬁmum extends over all countable covers from A. If µ(A) < ∞. For the σﬁnite case split again A as before. Let µ∗ be an outer measure. Theorem A. By regularity. This ﬁnishes the proof. Next let us turn to inner regularity. Hence K ˜ n ) → ∞ = µ(A). • A1 ⊆ A2 ⇒ µ∗ (A1 ) ≤ µ∗ (A2 ). Now we need to distinguish two cases: If µ(A) = ∞. Note that µ∗ (A) = µ(A) for A ∈ A (Problem A. there are compact subsets Kj of Aj such that µ(Aj ) ≤ µ(Kj ) + 2εj . Then O = j Oj is open.C sup µ(C) ≤ µ(A) compact (A.3). the sum µ(K j µ(Aj ) will converge and choosing n suﬃciently large we will have ˜ ˜ µ(Kn ) ≤ µ(A) ≤ µ(Kn ) + 2ε.
24) we have ∞ ∞ µ∗ (A) = k=1 µ∗ (Ak ∩ A) + µ∗ (A ∩ A) = k=1 µ∗ (Ak ) (A. Next. Without restriction we can assume that they are disjoint (compare the last argument in proof of ˜ Theorem A.2. . (A. (A. where we have used De Morgan and µ∗ (A ∩ B ∩ E) + µ∗ (A ∩ B ∩ E) + µ∗ (A ∩ B ∩ E) ≥ µ∗ ((A ∪ B) ∩ E) (A. let An be a sequence of sets from Σ. setting E = A in (A. Then for any set E we have ˜ ˜ ˜ µ∗ (An ∩ E) = µ∗ (An ∩ An ∩ E) + µ∗ (A ∩ An ∩ E) n (A.23) Letting n → ∞ and using subadditivity ﬁnally gives ∞ µ∗ (E) ≥ k=1 µ∗ (Ak ∩ E) + µ∗ (A ∩ E) (A. A = n An .20) ˜ = µ (An ∩ E) + µ (An−1 ∩ E) n ∗ ∗ = . B ∈ Σ. Proof. = k=1 µ∗ (Ak ∩ E).22) ˜ Using An ∈ Σ and monotonicity of µ∗ . we conclude that Σ is an algebra. Let A.A. Finally. Abbreviate An = k≤n An . Since the reverse inequality is just subadditivity. It clearly contains X and is closed under complements.24) ≥ µ∗ (A ∩ E) + µ∗ (B ∩ E) ≥ µ∗ (E) and we infer that Σ is a σalgebra. Applying Carath´odory’s condition e twice ﬁnally shows µ∗ (E) = µ∗ (A ∩ B ∩ E) + µ∗ (A ∩ B ∩ E) + µ∗ (A ∩ B ∩ E) + µ∗ (A ∩ B ∩ E) ≥ µ∗ ((A ∪ B) ∩ E) + µ∗ ((A ∪ B) ∩ E).3). Extending a premasure to a measure 217 (where A = X\A is the complement of A) form a σalgebra and µ∗ restricted to this σalgebra is a measure.21) which follows from subadditivity and (A ∪ B) ∩ E = (A ∩ B ∩ E) ∪ (A ∩ B ∩ E) ∪ (A ∩ B ∩ E)... We ﬁrst show that Σ is an algebra. we infer ˜ ˜ µ∗ (E) = µ∗ (An ∩ E) + µ∗ (An ∩ E) n ≥ k=1 µ∗ (Ak ∩ E) + µ∗ (A ∩ E).25) and we are done.
218
A. Almost everything about Lebesgue integration
Remark: The constructed measure µ is complete, that is, for any measurable set A of measure zero, any subset of A is again measurable (Problem A.4). The only remaining question is wether there are any nontrivial sets satisfying the Carath´odory condition. e Lemma A.8. Let µ be a premeasure on A and let µ∗ be the associated outer measure. Then every set in A satisﬁes the Carath´odory condition. e Proof. Let An ∈ A be a countable cover for E. Then for any A ∈ A we have
∞ ∞ ∞
µ(An ) =
n=1 n=1
µ(An ∩A)+
n=1
µ(An ∩A ) ≥ µ∗ (E∩A)+µ∗ (E∩A ) (A.26)
since An ∩ A ∈ A is a cover for E ∩ A and An ∩ A ∈ A is a cover for E ∩ A . Taking the inﬁmum we have µ∗ (E) ≥ µ∗ (E ∩ A) + µ∗ (E ∩ A ) which ﬁnishes the proof. Thus, as a consequence we obtain Theorem A.2. Problem A.2. Show that µ∗ deﬁned in (A.18) is an outer measure. (Hint for the last property: Take a cover {Bnk }∞ for An such that µ∗ (An ) = k=1 ∞ ε ∞ n An .) k=1 µ(Bnk ) and note that {Bnk }n,k=1 is a cover for 2n + Problem A.3. Show that µ∗ deﬁned in (A.18) extends µ. (Hint: For the cover An it is no restriction to assume An ∩ Am = ∅ and An ⊆ A.) Problem A.4. Show that the measure constructed in Theorem A.7 is complete.
A.3. Measurable functions
The Riemann integral works by splitting the x coordinate into small intervals and approximating f (x) on each interval by its minimum and maximum. The problem with this approach is that the diﬀerence between maximum and minimum will only tend to zero (as the intervals get smaller) if f (x) is suﬃciently nice. To avoid this problem we can force the diﬀerence to go to zero by considering, instead of an interval, the set of x for which f (x) lies between two given numbers a < b. Now we need the size of the set of these x, that is, the size of the preimage f −1 ((a, b)). For this to work, preimages of intervals must be measurable. A function f : X → Rn is called measurable if f −1 (A) ∈ Σ for every A ∈ Bn . A complexvalued function is called measurable if both its real and imaginary parts are. Clearly it suﬃces to check this condition for every
A.3. Measurable functions
219
set A in a collection of sets which generate Bn , say for all open intervals or even for open intervals which are products of (a, ∞): Lemma A.9. A function f : X → Rn is measurable if and only if f −1 (I) ∈ Σ
n
∀I =
j=1
(aj , ∞).
(A.27)
In particular, a function f : X → Rn is measurable if and only if every component is measurable. Proof. All you have to use is f −1 (Rn \A) = X\f −1 (A), f −1 ( j Aj ) = 1 j f (Aj ) and the fact that any open set is a countable union of open intervals. If Σ is the Borel σalgebra, we will call a measurable function also Borel function. Note that, in particular, Lemma A.10. Any continuous function is measurable and the composition of two measurable functions is again measurable. Moreover, sometimes it is also convenient to allow ±∞ as possible values for f , that is, functions f : X → R, R = R ∪ {−∞, ∞}. In this case A ⊆ R is called Borel if A ∩ R is. The set of all measurable functions forms an algebra. Lemma A.11. Suppose f, g : X → R are measurable functions. Then the sum f + g and the product f g is measurable. Proof. Note that addition and multiplication are continuous functions from R2 → R and hence the claim follows from the previous lemma. Moreover, the set of all measurable functions is closed under all important limiting operations. Lemma A.12. Suppose fn : X → R is a sequence of measurable functions, then inf fn , sup fn , lim inf fn , lim sup fn (A.28)
n∈N n∈N n→∞ n→∞
are measurable as well. Proof. It suﬃces to proof that sup fn is measurable since the rest follows from inf fn = − sup(−fn ), lim inf fn = supk inf n≥k fn , and lim sup fn = −1 inf k supn≥k fn . But (sup fn )−1 ((a, ∞)) = n fn ((a, ∞)) and we are done. It follows that if f and g are measurable functions, so are min(f, g), max(f, g), f  = max(f, −f ), f ± = max(±f, 0).
220
A. Almost everything about Lebesgue integration
A.4. The Lebesgue integral
Now we can deﬁne the integral for measurable functions as follows. A measurable function s : X → R is called simple if its range is ﬁnite s(X) = {αj }p , that is, if j=1
p
s=
j=1
αj χAj ,
Aj = s−1 (αj ) ∈ Σ.
(A.29)
Here χA is the characteristic function of A, that is, χA (x) = 1 if x ∈ A and χA (x) = 0 else.