See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/337733114

Sheet 2 of some exercices with detailed solutions on Functional Analysis and

Spectral Theory of linear operators.

Chapter · December 2019

CITATIONS READS

0 602

2 authors, including:

Abdelkader Intissar
Université de Corse Pascal Paoli
166 PUBLICATIONS 932 CITATIONS

SEE PROFILE

All content following this page was uploaded by Abdelkader Intissar on 04 December 2019.

The user has requested enhancement of the downloaded file.

 Exercises with detailed Solutions
On

Functional Analysis
And

Spectral Theory of Linear Operators

Sheet 2
Authors

Abdelkader INTISSAR and Jean-Karim INTISSAR

Functional Analysis Spectral Theory of Linear Operators

Sheet 2

• We begin this sheet by giving a characterization of an orthonormal basis in a separable Hilbert

space H.

Definition 14.2.1 (linear span of H1 ⊂ H)

Let H be a Hilbert space and H1 ⊂ H.

k
X
(1) The subset spanH1 := {ϕ ∈ H; ϕ = aj ϕj , ϕj ∈ H1 , aj ∈ C}
j=1

is called the linear span of H1 .

(2) A subset B ⊂ H is called a basis of H if B is linearly independent in H and spanB = H.

i.e. for every ϕ ∈ H and every  > 0 there exist k ∈ N and {aj }kj=1 ⊂ C such that

k
X
|| ϕ − aj ϕj ||< , ϕj ∈ B.
j=1

(3) The Hilbert space H is called separable if it has a countable or finite basis.

(4) An orthonormal system B = {ϕj } in H which is a basis is called an orthonormal basis.

By the Gram-Schmidt orthonormalization we may conclude that every separable Hilbert space
has an orthonormal basis.

FFF
Exercise 14.2.1 (characterization of an orthonormal basis)

Let B = {ϕj }∞
j=1 be an orthonormal system in a separable Hilbert space H.

Show that the following statements are equivalent :

(1) B is maximal i.e. it is not a proper subset of any other orthonormal system.

(2) For every ϕ ∈ H the condition < ϕ, ϕj >= 0, j = 1, 2, ... implies that ϕ = 0.
+∞
X
(3) Every ϕ ∈ H has the Fourier expansion ϕ = < ϕ, ϕj > ϕj
j=1

51

A. Intissar and J-K. Intissar

Functional Analysis Spectral Theory of Linear Operators

k
X
i.e. || ϕ − < ϕ, ϕj > ϕj ||−→ 0, as k −→ +∞. This means that B is an orthonormal basis.
j=1

(4) Every pair ϕ, ψ ∈ H satisfies the completeness relation

+∞
X
< ϕ, ψ >= < ϕ, ϕj > < ψ, ϕj >..
j=1

(5) Every ϕ ∈ H satisfies the Parseval equality

+∞
X
|| ϕ ||2 = |< ϕ, ϕj >|2 .
j=1

Solution 14.2.1

(1) =⇒ (2) Suppose that there is ϕ ∈ H, ϕ 6= 0 such that < ϕ, ϕj >= 0 for all j = 1, 2, ...
ϕ
Then B1 = { , ϕ1 , ϕ2 , ....} is an orthonormal system in H. This fact implies that B is not
|| ϕ ||
maximal. It contradicts (1) and proves (2).
k
X
(2) =⇒ (3) Given ϕ ∈ H introduce the sequence ϕ(k) = < ϕ, ϕj > ϕj . Theorem of Pytha-
j=1
k
X
goras and Bessel’s inequality imply that || ϕ(k) ||2 = |< ϕ, ϕj >|2 ≤|| ϕ ||2 .
j=1
k
X
It follows that |< ϕ, ϕj >|2 converges. That’s why,
j=1
k
X
(k) (m) 2
for m < k, || ϕ −ϕ || = |< ϕ, ϕj >|2 −→ 0 as k, m −→ +∞.
j=m+1

Hence ϕ(k) is a Cauchy sequence in H. Thus there exists ψ ∈ H such that

+∞
X
ψ = limϕ(k) = < ϕ, ϕj > ϕj as k −→ +∞.
j=1

Next, since the inner product is continuous we deduce that

< ψ, ϕj >= lim < ϕ(k) , ϕj >=< ϕ, ϕj > as k −→ +∞ for any j = 1, 2, ....

Therefore < ψ − ϕ, ϕj >= 0 for any j = 1, 2, .... Part (2) implies that ψ = ϕ and part (3) follows.
+∞
X +∞
X
(3) =⇒ (4) Let ϕ, ψ ∈ H. We know from part (3) that ϕ = < ϕ, ϕj > ϕj and ψ = < ψ, ϕj > ϕj .
j=1 j=1

Continuity of the inner product and orthonormality of {ϕj } allow us to conclude that
+∞ X
X +∞ +∞
X
< ϕ, ψ >= < ϕ, ϕj > < ψ, ϕk > < ϕj , ϕk >= < ϕ, ϕj > < ψ, ϕj >
j=1 k=1 j=1

(4) =⇒ (5) Take ψ = ϕ in part (4).

52

A. Intissar and J-K. Intissar

Functional Analysis Spectral Theory of Linear Operators

(5) =⇒ (1) Suppose that B is not maximal. Then we can add a unit vector ϕ ∈ H to it which
is orthogonal to B. Parseval’s equality gives then
+∞
X
1 =|| ϕ ||2 = |< ϕ, ϕj >|2 = 0. This contradiction proves the result.
j=1
F
Exercise 14.2.2

Consider the operator Af (x) = x−α f (x), α > 0 in the Hilbert space H = L2 [0, 1].

Let us define D(A) = {f ∈ L2 [0, 1]; f (x) = χn (x)g(x), g ∈ L2 f or some n ∈ N}

where

1

 0,
 0≤x≤
χn (x) = n
1
 1, <x≤1

n

(i) Show that D(A∗ ) = {v ∈ L2 [0, 1]; x−α v ∈ L2 [0, 1]}.

(ii) Show that A is not closed.

Solution 14.2.2

(i) It is clear that D(A) = L2 [0, 1]. For v ∈ D(A∗ ) we have

Z 1 Z 1
< Af, v >= x−α χn (x)g(x)v(x)dx = f (x)x−α v(x)dx =< f, A∗ v >.
0 0

That is why we may conclude that

D(A∗ ) = {v ∈ L2 [0, 1]; x−α v ∈ L2 [0, 1]}.

(ii) Let us show that A is not closed. To see this take the sequence :

1

 xα ,
 <x≤1
fn (x) = n
 1, 0 ≤ x ≤ 1

n

1

 1,
 <x≤1
Then fn ∈ D(A) and Afn (x) = n
1
 1, 0≤x≤

n
If we assume that A = A then

 fn ∈ D(A)




fn −→ xα




Afn −→ 1


53

A. Intissar and J-K. Intissar

 Functional Analysis Spectral Theory of Linear Operators

 α
 x ∈ D(A)
=⇒ . But xα ∈
/ D(A).
α
1 = Ax


This contradiction shows us that A is not closed. It is not bounded either since α > 0.

FF
Exercise 14.2.3

d2
Consider the operator A = in the Hilbert space H = L2 (0, 1) with domain :
dx2
D(A) = {f ∈ C 2 [0, 1]; f (0) = f (1) = f 0 (0) = f 0 (1) = 0}.

Show that the closure A of A is symmetric, but not self-adjoint.

Solution 14.2.3

It is clear that D(A) = L2 (0, 1) and A is not closed. Moreover, integration by parts gives
< Af, g >L2 =< f, Ag >L2 for any f ∈ D(A) and g ∈ W22 (0, 1). That is, A is symmetric such
that A ⊂ A∗ and D(A∗ ) = W22 (0, 1). As we know , A = A∗ always.
◦
Now we will show that A is the same differential operator of order 2 with D(A) =W22 (0, 1),
◦
where W22 (0, 1) denotes the closure of D(A) with respect to the norm of Sobolev space W22 (0, 1).

Indeed, for any f ∈ D(A) we have

|| Af ||2L2 + || f ||2L2 ≤|| f ||2W 2

2

and
Z 1
|| f ||2W 2 =|| Af ||2L2 + || f ||2L2 + | f 0 |2 dx
2
0
Z 1
3
=|| Af ||2L2 + || f ||2L2 − f f 00 dx ≤ (|| Af ||2L2 + || f ||2L2 .
0 2
◦
It means that || Af ||2L2 + || f ||2L2 ∼|| f ||2W 2 , gives now that D(A) = W22 (0, 1).
2

So we have finally
◦
D(A) ( D(A) = W22 (0, 1) = D(A∗∗ ) ( W22 (0, 1)

The closure A is symmetric, but not self-adjoint since

◦
∗
W22 (0, 1) = D(A) 6= D(A ) = D(A∗ ) = W22 (0, 1).

FF
Exercise 14.2.4

d2
Consider the operator A = in the Hilbert space H = L2 (0, 1) on R.
dx2

54

A. Intissar and J-K. Intissar

Functional Analysis Spectral Theory of Linear Operators

d2
(i) Show that A = with domain D(A) = C 1 [0, 1] is not symmetric.
dx2
d2
(ii) Show that A = 2 with domain D(A) = C01 [0, 1] = {u ∈ C 1 [0, 1]; u(0) = u(1) = 0} is sym-
dx
metric.

(iii) Show that the adjoint (A∗ , D(A∗ )) of the operator

d2
A= with domain D(A) = C01 [0, 1] = {u ∈ C 1 [0, 1]; u(0) = u(1) = 0}
dx2
is given by :

d2
A∗ = with domain D(A∗ ) = H0 = {v ∈ H, v 00 ∈ H, v(0) = v(1) = 0}
dx2
d2
(iv) Show that A = 2 with domain D(A) = C01 [0, 1] = {u ∈ C 1 [0, 1]; u(0) = u(1) = 0} is not
dx
self-adjoint

Solution 14.2.4

(i) Let u, v ∈ C 1 [0, 1] then

Z 1
(1) < Au, v >= u00 (x)v(x)dx
0
Z 1
= [u0 (x)v(x)]10 − u0 (x)v 0 (x)dx
0
Z 1
= [u0 (x)v(x)]10 − [u(x)v 0 (x)]10 + u(x)v 00 (x)dx
0 Z 1
0 0 0
= u (1)v(1) − u (0)v(0) − u(1)v (1) + u(0)v(0) + u(x)v 00 (x)dx
0
Z 1
(2) < u, Av >= u(x)v 00 (x)dx
0
Z 1
= [u(x)v 0 (x)]10 − u0 (x)v 0 (x)dx
0
Z 1
0 0 0
= [u(x)v (x)]10 − [u (x)v (x)]10 + u(x)v 00 (x)dx
0 Z 1
= u(1)v (1) − u(0)v (0) − u (1)v (1) + u0 (0)v 0 (0) +
0 0 0 0
u(x)v 00 (x)dx
0

Now by according (1) - (2), we get :

< u00 , v > − < u, v 00 >= u0 (1)v(1) − u(1)v 0 (1) + u(0)v 0 (0). (*)

and by choosing u(x) = x2 and v(x) = x we deduce that

d2
< u00 , v >= 1 and < u, v 00 >= 0 this implies that A = 2 with domain D(A) = C 1 [0, 1] is not
dx
symmetric.

55

A. Intissar and J-K. Intissar

Functional Analysis Spectral Theory of Linear Operators

d2
(ii) Let A = 2 with domain D(A) = C01 [0, 1] = {u ∈ C 1 [0, 1]; u(0) = u(1) = 0} then from (*)
dx
we deduce that

< u00 , v > − < u, v 00 >= u0 (1)v(1) − u(1)v 0 (1) + u(0)v 0 (0) = 0.∀ u ∈ C01 [0, 1], ∀ v ∈ C01 [0, 1] (**)

d2
this implies that A = with domain D(A) = C01 [0, 1] is symmetric.
dx2
d2
(iii) Let A = with domain D(A) = C01 [0, 1] then its adjoint (A∗ , D(A∗ )) is defined by
dx2
< Au, v >=< u, A∗ v >, u ∈ D(A) = C01 [0, 1], v ∈ D(A∗ where

D(A∗ ) = {v ∈ H; ∃ Cv > 0, |< Au, v >|≤ Cv || u || ∀ u ∈ D(A) = C01 [0, 1]

Let H0 be the space H0 = {v ∈ H, v 00 ∈ H, v(0) = v(1) = 0} and let v ∈ H0 then the mapping
` : C01 [0, 1] −→ R, u −→< u00 , v > is continuous. In fact :

As v 00 ∈ H = L2 [0, 1] then we have |< u00 , v >|≤|| u00 || . || v ||. By taking Cv =|| u00 || we deduce
that H0 ⊂ D(A∗ ).

Now, we will show that D(A∗ ) ⊂ H0 .

Let v ∈ D(A∗ ), then ∃ w ∈ H = L2 [0, 1] such that < u00 , v >=< u, w >, ∀ u ∈ D(A).

As < u00 , v >=< u, v 00 > this implies that v 00 = w ∈ H = L2 [0, 1].

Now as v ∈ L2 [0, 1] then v 0 ∈ L2 [0, 1] (a classical result) i.e., v belongs to Sobolev space
H 1 [0, 1] = {v ∈ L2 [0, 1], v 0 ∈ L2 [0, 1]}

Remark 14.2.1

(i) H 1 (Ω) in one dimension. Let n = 1 and Ω be a bounded interval. Then, each function from
H 1 (Ω) is continuous and bounded in Ω. In particular we get :
Z x p
∀ u ∈ H 1 ([0, 1]), we have u(x) = u(x0 ) + u0 (t)dt and | u(x) − u(y) |≤|| u0 || | x − y |.
x0

In fact :

The first equality is true almost everywhere and the inequality is true if we replace u (which
is a priori a function L2 [0, 1], thus defined only almost everywhere) by the expression given on
the left, which will therefore be its continuous representative.

Any function of H 1 ([0, 1]) is therefore continuous and 1/2-Höldérienne (or she has a Höldérian
representative, but it’s the same thing).

Any sequence (un ) bounded in H 1 ([0, 1]) is therefore equicontinuous (they are all Höderian
with the same exponent and the same constant) and any sequence that converges weakly in
H 1 ([0, 1]) converges uniformly thanks to Ascoli’s theorem. This is the same as saying that the
injection of H 1 ([0, 1]) in C([0, 1]) is compact.

56

A. Intissar and J-K. Intissar

Functional Analysis Spectral Theory of Linear Operators

Also, for all x0 ∈ [0, 1], we have :

p p
|| u ||H 1 ([0,1]) = || u ||L2 + || u0 ||L2 ≤ D || u ||L∞ + || u0 ||L2 ≤ C | u(x0 ) | + C || u0 ||L2

(ii) H 1 (Ω) in higher dimensions. The functions from H 1 (Ω) are in general not continuous for
n ≥ 2. This property will be shown with the following example :
1
Ω = {X ∈ R2 , || X ||2 ≤ } and u(X) = ln(| ln(|| X ||2 ) |) :
2

Graph of u(X)
.

Now as [0, 1] ⊂ R and the dimension of R is 1, then we deduce that

(α) v ∈ C 1 [0, 1].

Z 1
(β) < u00 , v >= u0 (1)v(1) − u0 (0)v(0) − u0 (x)v 0 (x)dx.
0

For this linear form to be continuous in v it is necessary that v(1) = v(0) = 0

In fact :

Let un nN be a sequence of functions that belongs to C 1 and satisfies :

 0

 | un (x) |≤ 1, x ∈ [0, 1]}




1

0
un (x) = 0, on [0, 1 − ]


 n


 0

un (1) = 1

By using the schwarz inequality and the mean value theorem we get
0 1
|| un ||≤|| un ||∞ ≤|| un ||≤ √ −→ 0 as n −→ +∞.
n
It follows that
Z 1
0 0
| un (x)v 0 (x)dx |≤|| un || . || v 0 ||−→ 0 as n −→ +∞.
0

and
Z 1
00 00
| un (x)v(x)dx |≤|| un || . || v ||−→ 0 as n −→ +∞.
0

57

A. Intissar and J-K. Intissar

Functional Analysis Spectral Theory of Linear Operators

Z 1
00 0
Since < un , v >= v(1) − un (x)v 0 (x)dx then v(1) = 0.
1
1− n
1
By taking the sequence of functions symmetric with respect the point 2 we deduce that v(0) = 0
and consequently D(A∗ ) = H0 .

(iv) As D(A) 6= H0 . then A is not self-adjoint

Remark 14.2.2

The spectral theory of two-point differential operators was begun by Birkhoff in his two papers
[ [ Birkhoff1] , [Birkhoff2 ]] of 1908, where he introduced regular boundary values for the
first time. It was continued by Stone [ [ Stone1] , [Stone2 ] with the initial work on irregu-
lar boundary values, and by Hoffman [Hoffman] in his thesis which examined second order
differential operators under irregular boundary values. An application of this spectral theory
to boundary problem with an operator in the boundary conditions or to abstract differential
equation are given in [Aliev-Yakubov], [Aimar et al] and in [Intissar] .

[ Birkhoff1] Birkhoff, G. D. :. Boundary value and expansion problems of ordinary linear dif-
ferential equations, Transactions of the American Mathematical Society, 9, (1908) 373-395

[ Birkhoff2] Birkhoff, G. D. :. On the asymptotic character of the solutions of certain linear

differential equations containing a parameter, Transactions of the American Mathematical So-
ciety, 9, (1908) 219-231

[ Stone1] Stone, M. H. :. A comparison of the series of Fourier and Birkhoff, Transactions of

the American Mathematical Society, 28, (1926) 695-761

[ Stone2] Stone, M. H. :. Irregular differential systems of order two and the related expansion
problems, Transactions of the American Mathematical Society, 29, (1927) 23-53

[ Hoffman] Hoffman, S. :. Second order linear differential operators defined by irregular boun-
dary conditions, Ph. D. thesis , Yale University, 1957.

[Aliev-Yakubov] Aliev, B. A. and Yakubov, S. Ya. :. A boundary - value problem with an

operator in boundary conditions, for a second order elliptic differential-operator equation, Sib
. Mat. Zh. , 26, No 4 (1990).

[Aimar et al] Aimar, M-T. , Intissar, A. and Paoli, J.M. :.Critères de complétude des vecteurs
propres généralisés d’une classe d’opérateurs no-auto-adjoints compacts ou à résolvante com-
pacte et applications. Publ. RIMS, Koyoto Univ., Vol. 32 , No 2, (1996)

[Intissar] Intissar, A. :. Analyse Fonctionnelle et Théorie Spectral pour les Opérateurs Com-
pacts Non Auto-Adjoints avec exercices et solutions. “Functional Analysis and Spectral Theory
for Non-Self-Adjoint Conpact Operators with Exercises and Solutions, Edition Cépaduès ,
(1997).

A complete spectral theory of non-self-adjoint two-point differential operators is given in [Lang-

Locker1], [Lang-Locker2] and by Locker in his excellent book [Locker]

[Lang-Locker1] Lang, P. and Locker, J. :. Spectral theory of two point differential operators

58

A. Intissar and J-K. Intissar

Functional Analysis Spectral Theory of Linear Operators

determined by −D2 . I. Spectral properties , Journal of Matimatical Analisis and Applications,

141, (1989)

[Lang-Locker2] Lang, P. and Locker, J. :. Denseness of the generalized eigenvectors of an H-S

discrete operator, Journal of functional Analysis , Vol. 82, No 2 , (1989).

[Locker] Locker, J. :. Spectral Theory of Non-Self-Adjoint Two-point Differential Operators,

American Mathematical Society, Vol. 73, (2000).

FFFFF
Exercise 14.2.5 (Lang-Locker)

d2
Let L = − 2 acting on H = L2 [0, 1] with domain D(L) = {u ∈ H 2 [0, 1]; B1 (u) = 0 and B2 (u) = 0}
dx
be the Locker’s operator.

where

• H 2 [0, 1] is the subspace of L2 [0, 1] consisting of all functions u ∈ C 1 [0, 1] such that u0 is abso-
lutely continuous on [0, 1] and u00 ∈ L2 [0, 1].

• B1 and B2 are linearly independent boundary values on H 2 [0, 1] given by

B1 (u) = a1 u0 (0) + b1 u0 (1) + a0 u(0) + b0 u(1) and B2 (u) = c1 u0 (0) + d1 u0 (1) + c0 u(0) + d0 u(1)

with the vectors (a0 , a1 , b0 , b1 ) and (c0 , c1 , d0 , d1 ) are linearly independent.

Give a complete spectral analysis of Locker’s operator.

Solution 14.2.5

A summary of the spectral properties of Locker’s operator (L, D(L) is given in following theo-
rem :

Theorem 14.2.3 (Lang-Locker , 1989)

(1) If a1 d1 − b1 c1 = 0 , a0 d0 − b0 c0 = 0 and a1 d0 − b0 c1 = a0 d1 − b1 c0 then the spectrum σ(L)

of (L, D(L) is σ(L) = ∅ or σ(L) = C.

(2) if σ(L) 6= ∅ and σ(L) 6= C then we have

(i) The resolvent of L is a Hilbert-Schmidt operator.

(ii) There exists a set of five rays γj : argλ = θj , j = 1, ..., 5 in the complexe plane such that
the angles between adjacent rays are less than π2 , for | λ | sufficiently large all the points on
1
the five rays belong to resolvent set ρ(L) of the operator L and || (λI − L)−1 ||= O( √ ) as
λ
λ −→ +∞ along each γj .

(iii) The system of generalized eigenfunctions of L is dense in H = L2 [0, 1].

59

A. Intissar and J-K. Intissar

Functional Analysis Spectral Theory of Linear Operators

Exercise 14.2.6FFFF (On the minimality and Riesz’s bases)

Z 1
Let H = L2 [0, 1], endowed of scalar product < u, v >= u(x)v(x)dx
0
Z 1
with its norm || u ||= | u(x) |2 dx.
0

We recall that :

• A system {ek (x)}∞

k=0 ∈ H is dense in H if, for u ∈ H such that < u, ek >= 0 ∀ k ∈ N then
u = 0.

• A system {ek (x)}∞

k=0 ∈ H is minimal in H if it is dense in H but it becomes non-dense in H
if we remove an element of this system.

• A system {ek (x)}∞

k=0 ∈ H is a basis in H if ∀ u ∈ H can be written uniquely in the following
form :
+∞
X
u= < u, ek > ek ( norm convergence).
k=0

• A system {ek (x)}∞

k=0 ∈ H is a Riesz’s basis in H if it is equivalent to orthonormal basis in H.

On H we consider the following three eigenvalue problems :

 00

 u (x) + µ2 u(x) = 0, 0 < x < 1



(P1 ) au(0) − u(1) = 0




au0 (0) + bu(1) + u0 (1) = 0; a ∈ C, b ∈ C, a 6= 0, ±1 and b 6= 0


 00

 u (x) + µ2 u(x) = 0, 0 < x < 1



(P2 ) u(0) = 0




 0
u (0) − u(1) = 0
 000

 y (x) = λy 0 (x), 0 < x < 1



 y 0 (0) + y(0) = 0



(P3 )
y 0 (1) = 0








 00
y (0) = 0

(1) Show that the systems of generalized functions associated to problems (P2 ) and (P2 ) res-
pectively are not the bases in H = L2 [0, 1].
Z x
(2) Let y(x) = −z(0) + z(s)ds, show that
0

60

A. Intissar and J-K. Intissar

Functional Analysis Spectral Theory of Linear Operators

• The problem (P3 ) is equivalent to

 00

 z (x) = λz(x), 0 < x < 1



(P4 ) z 0 (0) = 0




z(1) = 0


2 π
• The functions yk (x) = −1 + sin[ (2k + 1)x], k ∈ N are the eigenfunctions of the
π(2k + 1) 2
problem (P1 ).
p
(3) Let H1 = {u ∈ H; u0 ∈ H} equipped with its usual norm || u ||1 = || u ||2 + || u0 ||2 .
p
Show that || u ||? = | u(0) |2 + || u0 ||2 is a norm which is equivalent to || u ||1 .

(4) Show that the system {yk }∞

k=0 is dense in H1 and it is also dense in H.

(5) Show that the system {yk }∞

k=0 is minimal in H1 but it is not minimal in H.

(6) Show that the system {yk }∞

k=0 is a not Riesz basis in H1 and it is not a Riesz’s basis in H.

Solution 14.2.6

(1) Let L be a linear operator acting on L2 [0, 1] and it is defined by certain boundary conditions
and a differential expression `(u)(x) = u00 (x) + q(x) with q(x) is a complex potential belonging
to L1 [0, 1].

• A non-zero u0 (x) function is called an eigenfunction of L corresponding to the eigenvalue λ

0
if u0 (x) and u0 (x) are absolutely continuous and Lu0 + λu0 = 0.

• A non-zero up (x) function is called pth -eigenfunction associated to eigenfunction u0 (x) and to
0
eigenvalue λ if up (x) and up (x) are absolutely continuous and Lup + λup = up−1 , p = 1, 2, ....

• The system {up , p = 0, 1, ...} is called a system of generalized functions of the operator L.

• In [Gohberg-Krein], it was established that if a system {un , n ∈ N} of generalized eigen-

functions of L is dense and minimal in L2 [0, 1] then there exists an uniquely system {vn , n ∈ N}
satisfying :

< un , vk >= δnk

where

δnk = 0 if n 6= k and δnk = 1 if n = k.

Let {vn }n=1,2,.. the sequence of generalized eigenfunctions of the adjoint of L. We are in this
case if the boundary conditions are in Locker’s form. i,e.,

B1 (u) = a1 u0 (0) + b1 u0 (1) + a0 u(0) + b0 u(1) and B2 (u) = c1 u0 (0) + d1 u0 (1) + c0 u(0) + d0 u(1)

61

A. Intissar and J-K. Intissar

Functional Analysis Spectral Theory of Linear Operators

√
Let λn n = 1, 2, ... be the eigenvalues of L and βn their multiplicities and µn = λn then it is
well known that :

If there exists a subsequence {µnk } of µn satisfying the following properties :

• (i) | lim µnk |−→ +∞ as nk −→ +∞.

• (ii) βnk < +∞.

• (iii) =mµnk = o(<eµnk ) as k −→ +∞.

Then the system {un } of generalized functions associated to the operator L is not a basis in
L2 [0, 1].

ab
Now, for the problem P1 , we observe that µn = and it follows that the sequence µn
1 − a2
satisfies all above properties. So the system {un } of generalized functions associated to the
problem P1 is not a basis in L2 [0, 1].

For the problem P2 , we can consult [[Dunford-Schwartz], chapter XIX §, 6] where it is well
known that the system {un } of generalized functions associated to the problem P2 is not a basis
in L2 [0, 1].

In fact :

we observe that the eigenvalues of the problem P2 are of the form µn ∼ an + ibln(n) where
a ∈ R and b ∈ R, it follows that the above conditions • (i), • (ii) and • (iii) are saisfied.

(2) we consider the eigenvalue problem P3 :

 000

 y (x) = λy 0 (x), 0 < x < 1



 y 0 (0) + y(0) = 0



(P3 )
y 0 (1) = 0








 00
y (0) = 0
Z x
The substitution of y(x) = −z(0) + z(s)ds in P1 leads to the following equivalent eigenvalue
0
problem :
 00

 z (x) = λz(x), 0 < x < 1



(P4 ) z 0 (0) = 0




z(1) = 0


The eigenfunctions of the problem P4 are given by :

62

A. Intissar and J-K. Intissar

Functional Analysis Spectral Theory of Linear Operators

π
zk (x) = cos[ (2k + 1)x], k ∈ N.
2
Z x
As y(x) = −z(0) + z(s)ds then
0
2 π
• The functions yk (x) = −1 + sin[ (2k + 1)x], k ∈ N are the eigenfunctions of the
π(2k + 1) 2
problem (P1 ).

(3) see the remark of exercise 14.2.4 or by using the Schwarz’ s inequality as follows in the
following inequalities :
Z x
(3i) As u(x) = u(0) + u0 (t)dt =⇒
Z x 0
| u(x) |≤| u(0) | + | u0 (t) | dt =⇒
0
Z 1
| u(x) |≤| u(0) | + | u0 (t) | dt =⇒
0
s s
Z 1 Z 1
| u(x) |≤| u(0) | + 2
1 dt | u0 (t) |2 dt =⇒
0 0
0
| u(x) |≤| u(0) | + || u ||=|| u ||? =⇒

| u(x) |2 ≤|| u ||? =⇒

Z 1 Z 1
| u(x) |2 dx ≤|| u ||? dx =⇒
0 0

|| u ||2 ≤|| u ||2? =⇒

|| u ||≤|| u ||? =⇒

|| u || + || u0 ||≤ 2 || u ||? =⇒

|| u ||H 1 [0,1] ≤ 2 || u ||? .

Z x
(3ii) As u(0) = u(x) − u0 (t)dt =⇒
0
Z x
| u(0) |≤| u(x) | + | u0 (t) | dt =⇒
0
Z 1
| u(0) |≤| u(x) | + | u0 (t) | dt =⇒
0

| u(0) |≤| u(x) | + || u0 || =⇒

Z 1 Z 1 Z 1
| u(0) | dx ≤ | u(x) | dx + || u0 || dx =⇒
0 0 0
s s
Z 1 Z 1
| u(0) |≤ 12 dx | u(x) |2 dx+ || u0 || =⇒
0 0
0
| u(0) |≤|| u || + || u ||=|| u ||H 1 [0,1] =⇒

63

A. Intissar and J-K. Intissar

Functional Analysis Spectral Theory of Linear Operators

|| u ||? ≤ 2 || u ||H 1 [0,1] .

From (3i) and (3ii) we deduce that :

1
|| u ||H 1 [0,1] ≤|| u ||? ≤ 2 || u ||H 1 [0,1] .
2
(4) Let <, >? be the scalar product associated to the norm || . ||? and u ∈ H 1 [0, 1] such that
Z 1
< u, yk >? = 0 ∀ k ∈ N i.e. , < u, yk >? = u(0)yk (0) + u0 (x)yk (x)dx = 0 ∀ k ∈ N
0
Z 1
π
As yk (0) = −1, It follows that −u(0) + u0 (x)cos[ (2k + 1)x]dx = 0, ∀k ∈ N
0 2
These equations can be written in the following form :
Z 1
π
u(0) = u0 (x)cos[ (2k + 1)x]dx ∀ k ∈ N
0 2
Z 1
π
Then we deduce that u(0) = 0 in particular u0 (x)cos[ (2k + 1)x]dx = 0 ∀ k ∈ N
0 2
π
Now as the system {cos[ (2k + 1)x]}k∈N is dense in H = L2 [0, 1] , we deduce that u0 (x) = 0
2
and consequently u = 0

(5) We begin by showing that the system {yk }k∈N is not minimal in H = L2 [0, 1]. We suppose
that < u, yk >= 0, ∀ k ∈ N . Then
Z 1 Z 1
2 π
u(x)dx = u(x)sin[ (2k + 1)x]dx = ak , k = 1, 2, ....
0 π(2k + 1)
0 2
Z 1
As ak −→ 0 as k −→ +∞, we deduce that u(x)dx = 0 and ak = 0 , k = 0, 1, 2, ..., in particu-
0
π
lar u is orthogonal to the system {uk }k∈N where u0 (x) = 1 and uk (x) = sin[ (2k + 1)x], k = 1, 2, ....
2
π
S be the space generated by the system {sin[ (2k + 1)x], k = 1, 2, ....} which is orthogonal to
2
the space generated by the function {sin[ π2 x]}.
π
As < 1, sin[ x] >6= 0 then 1 ∈
/ S, so the system {yk }k∈N is not minimal in H = L2 [0, 1].
2
To show that the system {yk }k∈N is minimal in H 1 [0, 1], we consider the system {vn }n∈N defi-
4 π
ned by vn (x) = sin[ (2n + 1)x], n = 0, 1, 2, .....
π(2n + 1) 2
The systems {yk }k∈N and {vn }n∈N satisfy the following relation :

< yk , vn >? = δkn (δnk is Kronecker’s symbol). It follows that the system {vn }n∈N is minimal
in H 1 [0, 1] and the system {yk }k∈N is also minimal in H 1 [0, 1].

(6) It is obvious that the system {yk }k∈N is not a basis of Riesz in H = L2 [0, 1] because it is
not minimal in L2 [0, 1].

Now we suppose that the system {yk }k∈N is a basis of Riesz in H 1 [0, 1] then the adjoint system
{vn }n∈N is not dense in H 1 [0, 1] because vn (0) = 0, n ∈ N.

64

A. Intissar and J-K. Intissar

Functional Analysis Spectral Theory of Linear Operators

If the system {yk }k∈N was a basis of Riesz in H 1 [0, 1] then one would have < yk , u >? −→ 0 as
k −→ +∞ and for all u ∈ H 1 [0, 1]. But < yk , 1 >= −1 which proves that the system is not a
basis of Riesz in H 1 [0, 1].

Exercise 14.2.7FFFF ([Gohberg-Krein] see p. 320)

• Let V be a vector space (not necessarily finite dimensional).

A family of vectors {en }n∈I is a Hamel basis for V if

(a) the set of all finite linear combinations (finite linear span) of {en }n∈I is V.

(b) every finite subset of {en }n∈I is linearly independent.

We do not require the index set I of a Hamel basis to be countable.

In finite dimensional linear algebra, a Hamel basis is usually just called a “basis”. However, in
Banach spaces the term “basis” is usually reserved for Schauder basis :

• A sequence ({en }∞ n=1 in a Banach space X is called a Schauder basis (also called Countable
∞
X
basis) of X if for every u ∈ X there is a unique sequence of scalars {an }∞
n=1 so that u = an en .
n=1
A sequence {en }∞
n=1 which is a Schauder basis of its closed linear span is called a basic sequence.
Notice that the sequence {en }∞
n=1 is linearly independent.

Notice that a Hamel basis uses only finite dimensional linear combinations, while a Schauder
basis uses infinite ones.

• A total set (or fundamental set) in a normed space X is a subset M ⊂ X whose finite linear
span is dense in X.

65

A. Intissar and J-K. Intissar

Functional Analysis Spectral Theory of Linear Operators

An orthonormal basis is an orthonormal sequence in an inner product space X which is total in

X.

• Let H be a Hilbert space (or Banach) on K where K will denote the scalar field, either R or C
and let I be either N or {1, ..., m}. A sequence {un }n∈I of non-zero vectors is an unconditional
basic sequence in H if there is a constant M such that :
X X
|| n an un ||≤ M || an un || for every n , an ∈ K with | n |≤ 1.
n∈I n∈I

The best of such constants M is called the unconditional constant of {un }n∈I .

To clarify more the meaning of an unconditional basis we recall from [Lindenstrauss- Tza-
friri] (Proposition 1.c.6]) :

[Lindenstrauss-Tzafriri] Lindenstrauss, J. and Tzafriri, L. :.. Classical Banach Spaces I, Se-

quence Spaces. Springer-Verlag, Berlin, 1977.

the following result :

A basic sequence {un }∞

n=1 is unconditional if and only if any of the following conditions holds.

(i) For every permutation π of the integers the sequence {uπ(n) }∞

n=1 is a basic sequence.

∞
X
(ii) For every subset I of the integers the convergence of an un implies the convergence of
X n=1
an un .
n∈I
∞
X ∞
X
(iii) The convergence of an un implies the convergence of bn un whenever | bn |≤| an |,
n=1 n=1
for all n.

• We recall that the scalar products < u, v > and <<u, v>> are said to be topologically equi-
valent if theygenerate topologically equivalent norms, i.e. if there exist positive constants c1 , c2
such that c1 < u, u >≤ <<u, v>> ≤ c2 < u, v >.

• We recall that a basis {un }∞

n=0 for H is equivalent to a basis {vn } for H if there exists a
topological isomorphism S : H −→ H such that Sun = vn for all n.

• We recall that a sequence {un }∞

n=0 in a Hilbert space H is a Riesz basis for H if it is equivalent
to some (and therefore every) orthonormal basis for H.

• We recall that a sequence {un }∞

n=0 in a Hilbert space H is a Bessel sequence if

+∞
X
∀u ∈ H, |< u, un >|2 < +∞.
n=0

• We recall that a sequence {un }∞

n=0 in a Hilbert space H is :

(α) bounded below if inf || un ||> 0.

66

A. Intissar and J-K. Intissar

Functional Analysis Spectral Theory of Linear Operators

(β) bounded above if sup || un ||< ∞.

(γ) normalized if || un ||= 1.

• A short list of basic references on this topic :

[Gohberg-Krein] Gohberg, I. C. and Krein, M. G. :. Introduction to the theory of linear non-

selfadjoint operators, Volume 18, American Mathematical Society, Providence, 1969.

[Young] Young, R. :. An introduction to Nonhannonic Fourier Series, Academic Press, New

York, 1980.

[Marti] MARTI, J. :. Introduction to the Theory of Bases, Springer-Verlag, New York, 1969.

[Daubechies] Daubechies, I. :. Orthonormal bases of compactly supported wavelets, Comm.

Pure Appl. Math. 41 (1988), 909-996.

[Heil] Heil, C. E. :. Wiener Amalgam spaces in generalized Harmonic Analysis and Wavelet
Theory, Dissertation Phd , University of Maryland, 1990.

• Questions :

We will show in this exercise that the class of all bases that are equivalent to orthonormal bases
coincides with the class of all bounded unconditional bases for H, and we will discuss some of
the properties of such bases.

• We recall that a basis {un }∞ ∞

n=0 in a Hilbert space H is a bounded basis if {un }n=0 is norm-
bounded both above and below, i.e., if 0 < inf || un ||≤ sup || un ||< ∞.

(i) Show that all bases of Riesz are preserved by topological isomorphisms. That is, if {un }∞
n=0
is a Riesz basis for a Hilbert space H and S : H −→ H̃ is a topological isomorphism, then
{Sun} is a Riesz basis for the Hilbert space H̃.

(ii) Show that all Riesz bases are bounded unconditional bases.

(iii) Let (({un }, {vn })) be a basis for a Hilbert space H where {un } possesses the biorthogonal
sequence {vn }. Then the following statements are equivalent.

(µ) {un } is a Riesz basis for H.

(λ) {vn } is a Riesz basis for H.

(δ) {un } ∼ {vn }.

(iv) Show that if a sequence {un }∞ n=0 is a Bessel sequence, then the coefficient mapping
C : H −→ `2 , u −→ C(u) = (< u, un >) is is a continuous linear mapping of H into `2 .
In other words, there exists a constant B > 0 such that :
+∞
X
∀u ∈ H, |< u, un >|2 ≤ B || u ||.
n=0

67

A. Intissar and J-K. Intissar

Functional Analysis Spectral Theory of Linear Operators

(v) Show that If {un } be a sequence in a Hilbert space H, then the following statements are
equivalent.

(a) {un } is a Riesz basis for H.

(b) {un } is a bounded unconditional basis for H.

∞
X ∞
X
(c) {un } is a basis for H , and cn un converges =⇒ | cn |2 < ∞.
n=0 n=0

(d) {un } is complete in H and there exist constants A, B > 0 such that
N
X N
X N
X
∀ c1 , ..., cN , A | cn |2 ≤|| cn un ||≤ | cn |2
n=0 n=0 n=0

(e) There is an equivalent inner product ((ů, ů)) for H such that {un } is an orthonormal basis
for H with respect to ((ů, ů)).

(f ) {un } is a complete Bessel sequence and possesses a bi-orthogonal system {vn } that is also
a complete Bessel sequence

Solution 14.2.7

(i) Since H possesses a basis, it is separable. Therefore H̃, being isomorphic to H, is separable
as well. It is well known that all separable Hilbert spaces are isometrically isomorphic, so there
exists an isometry U that maps H onto H̃. Further, by definition of Riesz basis, there exists
an orthonormal basis {en }∞ n=0 for H and a topological isomorphism T : H −→ H such that
T en = un .

Since U is an isometric isomorphism, the sequence {U (en )}∞ n=0 is an orthonormal basis for
H̃. Hence, ST U −1 is a topological isomorphism of H̃ onto itself which has the property that
ST U −1 (U (en )) = ST (en ) = S(un ). Hence {S(en )} is equivalent to an orthonormal basis for H̃,
so we conclude that {S(en )} is a Riesz basis for H̃.

(ii) Let {un }∞

n=0 be a Riesz basis for a Hilbert space H. Then there exists an orthonormal basis
{en }∞
n=0 for H and a topological isomorphism S : H −→ H such that Sen = un for every n.
However, {en }∞n=0 is a bounded unconditional basis, and bounded unconditional bases are pre-
served by topological isomorphisms then {un }∞ n=0 must be a bounded unconditional basis for H.

(iii)
(µ)=⇒(λ); (δ).

We begin by recalling that :

(•) Let {an } be a bi-orthogonal to {un } and {bn } be bi-orthogonal to {vn } and let ({un }, {an })
and ({vn }, {bn }) be bases for a Hilbert space H. If {un } ∼ {vn } then {an } ∼ {bn } (?).

(••) Let {vn } bi a bi-orthogonal to {un } in a Hilbert H, then ({un }, {vn }) is a basis, uncondi-
tional basis, or bounded basis for H ⇐⇒ the same is true of ({vn }, {un }) (??).

Now, assume that {un } is a Riesz basis for H. Then {un } ∼ {en } for some orthonormal basis
{en } of H. By (?), it follows that {un } and {en } have equivalent bi-orthogonal systems. Howe-

68

A. Intissar and J-K. Intissar

Functional Analysis Spectral Theory of Linear Operators

ver, {en } is bi-orthogonal to itself, so this implies

{vn } ∼ {en } ∼ {un }.

Hence {vn } is equivalent to {un }, and {vn } is a Riesz basis for H.

(λ)=⇒(µ); (δ).

By (??) ({vn }, {un }) is a basis for H. Therefore, this argument follows symmetrically.

(δ)=⇒(µ); (λ).

Assume that {un } ∼ {vn }. Then there exists a topological isomorphism S : H −→ H such that
Sun = vn for every n. Since (({un }, {vn }) is a basis, it follows that for each u ∈ H,
X X X
u= < u, vn > un = < u, Sun > un whence Su = < u, Sun > Sun
n n n
X
Therefore, < Su, u >= |< u, Sun >|2 ≥ 0.
n

Thus S is a continuous and positive linear operator on H, and therefore has a continuous and
1
positive square root S 2 [Wei80, Theorem 7.20]. Similarly, S −1 is positive and has a positive
1 1 1
square root, which must be S − 2 = [S 2 ]−1 . Thus S 2 is a topological isomorphism of H onto
1
itself. Moreover, S 2 is self-adjoint, so
1 1 1 1
< S 2 um , S 2 un >=< um , S 2 S 2 un >=< um , Sun >=< um , vn >= δmn
1
Hence {S 2 un } is an orthonormal sequence in H, and it must be complete since {un } is com-
1
plete and S 2 is a topological isomorphism. Therefore {un } is the image of the orthonormal basis
1 1
{S 2 un } under the topological isomorphism S − 2 . Hence {un } is a Riesz basis. By symmetry,
{vn } is a Riesz basis as well.

(iv) We will use the Closed Graph Theorem to show that C is continuous. Suppose that
vN −→ v ∈ H, and that C(uN ) −→ (cn ) ∈ `2 . Then for each fixed m
s
X
| cm − < vN , um >|≤ | cm − < vN , um >| =|| (cn − CvN ) ||`2 −→ 0 as N −→ +∞.
n

Therefore cm = lim < vN , um > as N −→ +∞ for every m. Hence (cn ) = (< v, um >) = C(v),
so C has a closed graph, and therefore is continuous.

The constant B is sometimes referred to as a Bessel bound or upper frame bound for {un }∞
n=0 .

(v) (a)=⇒(b) This is the content of (ii)

(a) ⇐⇒ (c). Assume that {un } is a basis for H, and let {en } be any orthonormal basis for H.

Then {un } is a Riesz basis for H if and only if {un } ∼ {en } By the following theorem [Young] :

69

A. Intissar and J-K. Intissar

Functional Analysis Spectral Theory of Linear Operators

Theorem 14.2.4 [Young] p. 30)

Let X and Y be Banach spaces. Let {xn } be a basis for X and let {yn } be a basis for Y . Then
the following two statements are equivalent.

(•1 ) {xn } is equivalent to {yn }.

X X
(•2 ) cn xn converges in X if and only if cn yn converges in Y.
n n

It follows that
∞
X ∞
X
{un } ∼ {en } if and only if cn un converges ⇐⇒ cn en converges .
n=0 n=0

However, by the following Theorem :

Theorem 14.2.5

Let {xn } be an orthonormal sequence in a Hilbert space H.

X
(α) The series x = cn un converges ⇐⇒ (cn ∈ `2 . In this case we have the Plancherel
X n
Formula || u ||2 = | cn |2 .
n
X
(β) If x = cn xn converges then cn =< x, xn >. In particular, (cn ) = (< x, xn >) is the
n X
unique choice of coefficients such that x = cn xn .
n
X
(γ) (Bessel Inequality) If x ∈ H then | cn |2 ≤|| x ||2 .
n

(δ) the following statements are equivalent :

(i) {xn } is complete in H.

(ii) {xn } is an orthonormal basis for H.

X
(iii) (Plancherel Formula) |< x, xn >|2 =|| x ||2 for every x ∈ H.
n
X
(iv) < x, xn > xn = x for every x ∈ H.
n

It follows that
∞
X ∞
X
cn un converges ⇐⇒ | cn |2 < ∞. Hence, statement (a) holds if and only if statement
n=0 n=0
(c) holds.

(a)=⇒(d). Suppose that {un } is a Riesz basis for H, Then there exists an orthonormal basis
{en } for H and a topological isomorphism S : H −→ H such that Sen = un for every n. The-
refore, for any scalars c1 , ....., cN we have

70

A. Intissar and J-K. Intissar

Functional Analysis Spectral Theory of Linear Operators

N
X N
X N
X N
X
|| cn un ||2 =|| S( cn un ) ||2 ≤|| S ||2 || cn un ||2 =|| S ||2 | cn |2 ;
n=0 n=0 n=0 n=0

the last equality following from the Plancherel formula (se above Theorem). Similarly
N
X N
X N
X N
X
| cn |2 =|| cn en ||2 =|| S −1 ( cn un ) ||2 ≤|| S −1 ||2 || cn un ||2
n=0 n=0 n=0 n=0

Hence statement (d) holds with A =|| S −1 ||2 and B =|| S ||2 .

(a)=⇒(e). Suppose that {un } is a Riesz basis for H. Then there exists an orthonormal basis
{en } for H and a topological isomorphism S : H −→ H such that Sen = un for every n. Define

<< u, v >>=< Su; Sv > and ||| u |||2 =<< u, u >>=< Su, Su >=|| Su ||2

It is easy to see that <<, >> is an inner product for H, and that ||| . ||| is the corresponding
induced norm. Further,

||| u |||2 =|| Su ||2 ≤|| S ||2 . || u ||2 and || u ||2 =||| S −1 u |||2 ≤||| S −1 |||2 . ||| u |||2 (*)

where ||| S −1 ||| is the operator norm of S −1 with respect to the norm ||| . |||. In fact, we have

||| S −1 |||= sup|||u|||=1 =||| S −1 u |||= sup||Su||=1 || u ||= sup||v||=1 || S −1 v ||=|| S −1 ||.

It follows from (*) that ||| . ||| is an equivalent norm to || . ||. By definition , <<, >> is therefore
an equivalent inner product to <, >.

It remains to show that {un } is an orthonormal basis with respect to the inner product <<, >>.
By using the above Theorem 14.2.5

it suffices to show that {un } is a complete orthonormal sequence with respect to <<, >>. The
orthonormality follows from the calculation :

<< um , un >>=<< Sum , Sun >=< um , un >= δmn .

For the completeness, suppose that there is an u ∈ H such that << u, un >>= 0 for every
n. Then 0 =<< u, un >>=< Su, Sun >=< Su, en > for every n. Since {en } is complete with
respect to <, > this implies that Su = 0. Since S is a topological isomorphism, we therefore
have u = 0. Hence {un } is complete with respect to <<, >>.

(a)=⇒(f ). Suppose that {un } is a Riesz basis for H. Then, by (iii), {un } possesses a bi-
orthogonal sequence {vn } which is itself a Riesz basis for H.
X
Suppose now that u ∈ H. Then since ({un }; {vn }) is a basis, we have that u = < u, vn > un .
n

X that statement (a) implies statement (c), the convergence of this

Since we have already shown
series implies that u = |< u, vn >|2 < ∞.
n

Therefore {vn } is a Bessel sequence. Further, {vn } is complete since ({vn }; {un }) is also a ba-
sis for H ((??)). A symmetric argument implies that {un } is a complete Bessel sequence as well.

71

A. Intissar and J-K. Intissar

Functional Analysis Spectral Theory of Linear Operators

(b)=⇒(f ). Suppose that ({un }; {vn }) is a bounded unconditional basis for H. Then, by ((??)),
({vn }; {un }) is also a bounded unconditional basis for H. Therefore, if u ∈ H then
X
u= < u, un > vn , with unconditional convergence of this series.
n

By Orlicz’s Theorem (Theorem 3.1 in [Orlicz]) :

[Orlicz] Orlicz, W. :. Uber unbedingte Konvergence in Funtionenraumen, I, Studia Math. 4

(1933), 33-37.
X
we deduce that |< u, un >|2 || vn ||2 < ∞. However, by definition of bounded basis, there
n X
exist constants C1 , C2 so that 0 < C1 ≤|| vn || C2 < ∞ for all n. Hence |< u, un >|< ∞, so
n
{un } is a Bessel sequence, and it must be complete since it is a basis. A symmetric argument
implies that {vn } is also a complete Bessel sequence.

(d)=⇒(a).. Suppose that statement (d) holds, and let {en } be any orthonormal basis for H.
X
Choose any u ∈ H. Then, by (property δ of theorem 14.2.5), u = < u, en > en , and |< u, en |2 =|| u ||2 < ∞.
n

Choose M < N , and define c1 = .... = cM = 0 and cn =< u, en > for n = M + 1, ...., N . Then,
by hypothesis (d) we deduce that :
N
X N
X N
X N
X
|| < u, en > un ||2 =|| cn un ||2 ≤ B | cn |2 = B |< u, en >|2 .
n=M +1 n=1 n=1 n=1
2
Since |< u, en >| is a Cauchy series of real numbers, it follows that < u, en >
X un is a Cauchy
series in H and hence must converge in H. Therefore, we can define Su = < u, en > un .
n
Clearly S defined in this way is a linear mapping of H into itself, and we claim that S is a
topological isomorphism of H onto itself.

By applying hypothesis (d) and taking the limit as N −→ +∞, we have

X X
A || u ||2 = A |< u, en >|2 ≤|| Su ||2 ≤ B |< u, en >|2 = B || u ||2 . (**)
n n

It follows that S is continuous and injective, and that S −1 : Range(S) −→ H is continuous as

well.
X
Further, Sem = < em ; en > un = un for every n, so Range(S) contains every um , and there-
n
fore contains span({un }, which is dense in H since ({un } is complete. Therefore, if we show that
Range(S) is closed, then it will follow that Range(S) = H and hence that S is a topological
isomorphism of H onto itself.

Suppose then that vn ∈ Range(S) and that vn −→ v ∈ H. Then there exist un ∈ H such that
Sun = vn . Hence {Sun } is a Cauchy sequence in H. However, by (**) we have

A || um − un ||≤|| Sum − Sun ||,

72

A. Intissar and J-K. Intissar

Functional Analysis Spectral Theory of Linear Operators

so {un } is Cauchy as well, and therefore must converge to some u ∈ H. Since S is continuous, it
follows that vn = Sun −→ Su. Since we also have vn −→ v, we must have v = Su ∈ Range(S).
Hence Range(S) is closed.

Thus S is a topological isomorphism of H onto itself. Finally, since S maps {en } onto {un }, we
conclude that {un } is a Riesz basis for H.

(d))=⇒(b). Suppose that statement (d) holds. Choose N > 0, and define an = δN n . Then, by
hypothesis (d),
XN N
X N
X
A=A | an |2 ≤|| | an un ||2 =|| uN ||2 ≤ B | an |2 = B.
n=1 n=1 n=1

Hence {un } is norm-bounded above and below. In particular, each un is nonzero.

It remains to show that{un } is an unconditional basis. Therefore, choose any scalars c1 , ......, cN
and any signs 1 , ...., N = ±1. Then by hypothesis (d),
N N N N
X X X B X
|| | n cn un ||2 ≤ B | n cn |2 = B | cn | 2 ≤ || | n cn un ||2
n=1 n=1 n=1
A n=1
This, combined with the fact that {un } is complete and that every un is nonzero, implies that
{un } is an unconditional basis for H by applying the following theorem :

Theorem 14.2.6 (see [Singer], p. 461)

Let {xn } be a complete sequence in a Banach space X such that xn 6= 0 for every n. Then the
following statements are equivalent.

(1) {xn } is an unconditional basis for X.

(2) ∃ C1 ≥ 1, ∀ c1 , ......, cN , ∀ 1 , ...., N = ±1,

N
X N
X
|| | n cn xn ||≤ C1 || | cn xn ||.
n=1 n=1

(3) ∃ C2 ≥ 1, ∀ b1 , ......, bN , ∀ c1 , ......, cN , | b1 |≤| c1 |, ........ | bN |≤| cN |,

N
X N
X
=⇒|| | bn | xn ||≤ C2 || | cn | xn ||.
n=1 n=1

(4) ∃ 0 < C3 ≤ 1 ≤ C4 < ∞, ∀ c1 , ......, cN ,

N
X N
X N
X
C3 || | cn | xn ||≤|| cn xn ||≤ C4 || | cn | xn ||.
n=1 n=1 n=1

(5) {xn } is a basis, and for each bounded sequence of scalars (λn ) there exists a continuous
linear operator T : X −→ X such that T(xn ) = λn xn for all n.

[Singer] Singer, I. :. Bases in Banach Spaces I, SpringerVerlag, New York, 1970.

(d)=⇒(c).. Suppose that statement (d) holds. Choose N > 0, and define an = δN n . Then by
hypothesis (d),

73

A. Intissar and J-K. Intissar

Functional Analysis Spectral Theory of Linear Operators

N
X N
X
|| uN ||2 =|| an un ||2 ≥ | an |2 = A.
n=1 n=1

Hence {un } is norm-bounded below. In particular, each un is nonzero.

We will show now that {un } is a basis for H. To do this, choose any M < N , and any scalars
c1 , ...., cN . Then, by hypothesis (d),
M M N N
X X X B X
|| cn un ||2 ≥ B | cn |2 ≤ B | cn |2 ≤ || cn un ||2 .
n=1 n=1 n=1
A n=1

This, combined with the fact that {un } is complete and that every un is nonzero, implies that
{un } is a basis by applying the following theorem :

Theorem 14.2.7 (see [Singer], p. 25)

Let {xn } be a sequence in a Banach space X. Then the following statements are equivalent.

(1) {xn } is a basis for X.

X
(2) There exists a bi-orthogonal sequence {an } ⊂ X∗ such that ∀ x ∈ X, x = < x, an > xn .
n

(3) {xn } is complete and there exists a bi-orthogonal sequence {an } ⊂ X∗ such that supN || SN x ||< ∞.

(4) {xn } is complete and there exists a bi-orthogonal sequence {an } ⊂ X∗ such that supN || SN ||< ∞.
X X
It therefore only remains to show that cn un converges if and only if | cn |2 < ∞.
n n

To do this, let (cn ) be any sequence of scalars. Choose any M < N , and define

a1 = ..... = aM = 0 and an = cn for n = M + 1, ....., N . Then, by hypothesis (d),

N
X N
X N
X
A | an |2 ≥|| an un ||2 =≤ B | an |2 .
n=1 n=1 n=1

However, by the definition of an , this simply states that

N
X N
X N
X
A | cn |2 ≥|| cn un ||2 =≤ B | cn |2 .
n=1 n=1 n=1
X N
X
Therefore, cn un is a Cauchy series in H if and only if | cn |2 is a Cauchy series of real
n n=1
numbers.

Hence one series converges if and only if the other series converges.

(e)=⇒(d). Suppose that <<, >> is an equivalent inner product for H such that {un } is an
orthonormal basis with respect to <<, >>. Let ||| . ||| denote the norm induced by <<, >>.
Then, by definition of equivalent inner product, ||| . ||| and || . || are equivalent norms, i.e., there
exist constants A, B > 0 such that

74

A. Intissar and J-K. Intissar

Functional Analysis Spectral Theory of Linear Operators

∀ u ∈ H; A ||| u |||2 ≤|| u ||2 ≤ B ||| u |||2 (***)

Since {un } is complete in the norm ||| . ||| and since || . || is equivalent to ||| . |||, we must have
that {un } is complete in H with respect to || . ||. To see this explicitly, suppose that u ∈ H and
that  > 0 is given. Then since span {un } is dense in H with respect to the norm ||| . |||, there
must exist v ∈ span{un } such that ||| u − v |||< .
1
By (***), we therefore have || u − v ||< B 2 . Hence span{un } is also dense in H with respect
to || . ||, and therefore {un } is complete with respect to this norm.

Now choose any scalars c1 , ........, cN . Since {un } is orthonormal with respect to <<, >>, we
have by the Plancherel formula of above theorem that
N
X N
X
||| cn un |||= | cn |2 . Combined with (***), this implies that
n=1 n=1
N
X N
X N
X
A | cn |2 ≤|| cn un ||2 ≤ B | cn |2 .
n=1 n=1 n=1

Hence statement (d) holds.

(f )=⇒(c)

Suppose that statement (f ) holds. Since {un } and {vn } are both Bessel sequences, it follows
from the question (iv) that there exist constants C, D > 0 such that
X X
∀u ∈ H, |< u, un >|2 ≤ C || u ||2 and |< u, vn >|2 ≤ C || u ||2 (****)
n n

We will show now that ({un } ; {vn }) is a basis for H. Since {un } is assumed to be complete and
since {vn } is bi-orthogonal to {un }, it suffices by above Theorem to show that sup || SN ||< ∞,
where SN is the partial sum operator
N
X
SN u = < u, vn > un .
n=1

We compute :

|| SN ||2 = sup||v||=1 |< SN u, v >|2 ( by using || u ||= sup||v||=1 |< u, v >|)

N
X
= sup||v||=1 | < u, vn >< un , v >|2
n=1
N
X N
X
≤ sup||v||=1 ( |< u, vn >|2 )( |< un , v >|2 ) (by CauchySchwarz)
n=1 n=1

≤ sup||v||=1 D || u ||2 C || v ||2 (****)

= CD || u ||2

Hence sup || SN ||2 ≤ CD < ∞, as desired.

75

A. Intissar and J-K. Intissar

Functional Analysis Spectral Theory of Linear Operators

X X
Finally, we must show that cn un converges if and only if | cn |2 < ∞.
n n
X
Suppose firstu = cn un converges. Then we must have cn =< u, vn > since({un }, {vn }) is a
n X X
basis for H. It therefore follows from (****) that | cn |2 = |< u, vn >|2 ≤ D || u ||2 .
n n
X
Conversely, suppose that | cn |2 < ∞. Then for any M < N ,
n
X N
X
|| cn un ||2 = sup||v||=1 |< cn un , v >|2 ( by using || u ||= sup||v||=1 |< u, v >|)
n n=M +1
N
X
= sup||v||=1 |< cn < un , v >|2
n=M +1
N
X N
X
= sup||v||=1 ( | cn |2 )( |< un , v >|2 ) by Cauchy Schwarz
n=M +1 n=M +1
XN
≤ sup||v||=1 ( | cn |2 )C || v ||2 by (****)
n=M +1
N
X
= | cn |2 .
n=M +1
X
Hence cn un is a Cauchy series in H, and therefore must converge.
n

In view of the absence of recent textbooks in which one might find a discussion of the basic
theory of bases in a Hilbert space, we recommend the excellent chapter VI of Gohberg-Krein’s
book on this topic in particular the theorem of Bari on the number of characteristic properties
of bases equivalent to orthonormal bases :

Theorem 14.2.8 (Bari’s theorem [Bari] proved in [Gohberg-Krein, p. 310])

The following assertions are equivalent.

(1) The sequence {u}∞

n forms a basis of the Hilbert space H, equivalent to an orthonormal basis.

(2) The sequence {u}∞ n becomes an orthonormal basis of the space H following the appropriate
replacement of the scalar product < u, v > by some new one <<u, v>>, topologically equiva-
lent to the original one.

(3) The sequence {u}∞ n is complete in H, and there exist positive constants A, B such that for
any positive integer N and any complex numbers c1 , c2 , ......, cN one has
N
X N
X N
X
A | cn |2 ≤| | cn un |2 ≤ B | cn |2 .
n=1 n=1 n=1

(4) The sequence {u}∞

is complete in H, and its Gram matrix generates a bounded invertible
n
operator in the space `2 .

(5) The sequence {u}∞ n is complete in H, there corresponds to it a complete bi-orthogonal

sequence {v}∞
n , and for any u ∈ H one has

76

A. Intissar and J-K. Intissar

Functional Analysis Spectral Theory of Linear Operators

∞
X ∞
X
|< u, un >|2 < ∞ and |< u, vn >|2 < ∞ .
n=1 n=1

[Bari] Bari, N. K. :. Bi-orthogonal systems and bases in Hilbert space, Moskov, Gos. Univ.
Ucem. Zap. 148, Matematika 4 (1951) 69-107.

Remark 14.2.9

• The equivalence of assertion (1) and (4) was proved independently by R.P. Boas in

[Boas] Boas, R. P. :. A general moment problem , Amer. J ; Math. 63 (1941) 361-370

• This exercice was given as a theorem with is proof in volume 2 page 320 of book’s Intissar :

[Intissar] Intissar, A. and Intissar, J. K. :. Spectral Theory for Non-Self-adjoint operators with
compact resolvent and applications to Gribov-Intissar and other operators, Tome 2, Chapter
IV and V, Kindle Edition, (2019)

Exercise 14.2.8FFFF

L. L. Littlejohn and R. Wellman. A general left-definite theory for certain self-adjoint operators
with applications to differential equations. J. Differential Equations,181(2) :280 ?339, 2002.

We consider the differential equation :

a2 (x)y 00 (x) + a1 (x)y 0 (x) + (a0 (x) − λ)y(x) = 0, x ∈ [a, b], (1)

where a0 , al and a2 are functions of x such that when (1) is written in the equivalent form :

(py 0 )0 + (q − λ)y = 0, x ∈ [a, b], (2)

with p = wa2 , p0 = wa1 and q = wao ,

(i) p, p0 , q and w are real valued and continuous for x ∈ [a, b],

(ii) p(x)> 0, q(x)> 0 for x ∈ [a, b].

Equation (1) or (2), with boundary conditions :

B1 (y) = α1 y(a) + β1 y 0 (a) + γ1 y(b) + δ1 y 0 (b) = 0





(3)


B2 (y) = α2 y(a) + β2 y 0 (a) + γ2 y(b) + δ2 y 0 (b) = 0


where
 
α1 β1 γ1 δ1
  (4)
α2 β2 γ2 δ2

77

A. Intissar and J-K. Intissar

Functional Analysis Spectral Theory of Linear Operators

is real and of rank 2, defines (see [ Hellwig]) a linear operator L acting on D by

1


 (i) L(y) = [(py 0 )0 + qy], ∀ y ∈ D



 w



(ii) D = {y ∈ C 2 [a, b]; B1 (y) = 0, B2 (y) = 0} (5)



 Z b Z b


 (iii) D ⊂ H := L2 ([a, b], wdx) = {y; 2
| y | wdx < +∞}, < y1 , y2 >= y1 y2 wdx


a a

[ Hellwig] G. Hellwig, Differenlial Operators of Mathemaiical Physics, Addison Wesley Publi-

shing CO., hc.,Fteading, Massachusetts, (1967).

We consider the following transform defined by :

 
f (a)
 f 0 (a) 
T : C 1 [a, b] −→ R4 , f −→ T (f ) = 
 f (b) 


f 0 (b)
We use this mapping to associate with each eigenfunction y of L with domain D a pair of points
U ∈ R4 , V ∈ R4 defined as follow :

For y(x) = u(x) + iv(x), i2 = −1 is associated Y = U + iV where

   
u(a) v(a)
 u0 (a)   v 0 (a) 
U =  u(b)  = T (u) and V =  v(b)  = T (v)
  

u0 (b) v 0 (b)
  
αi u(a) + iv(a)
 βi   u0 (a) + iv 0 (a) 
Now let Bi = 
 γi , i = 1, 2 and Y =  u(b) + iv(b)
   . Then we have :

δi u0 (b) + iv 0 (b)

Bi (y) = 0 ⇐⇒ Y T Bi = 0, i = 1, 2.

Definition 14.2.10

L∗ with domain D∗ is said to be adjoint to L with domain D if

< Ly1 , y2 >=< y1 , L∗ y2 > for all y1 ∈ D, y2 ∈ D∗ .

(i) Show that if L is specified by (i) of (5), we have

0 0
< Ly1 , y2 > − < y1 , Ly2 >= [p(y1 y2 − y1 y2 )]ba (6)

(ii) Let {B3 , B4 } be a spanning set for << B1 , B2 >>T , show that the adjoint boundary condi-
tions for D∗ are specified by the conditions :

B1∗ (y2 ) = 0, B2∗ (y2 ) = 0 (7)

78

A. Intissar and J-K. Intissar

Functional Analysis Spectral Theory of Linear Operators

where

B1∗ = M B3 and B2∗ = M B4 with

 
0 −p(a) 0 0
 
 
 p(a) 0 0 0 
 
M= 


 0 0 0 p(b) 
 
 
0 0 −p(b) 0

(iii) Show that the matrix M is nonsingular and it is antisymmetric.

(iv) What conditions on D∗ such that L is self-adjoint ?

(v) Let u1 = u1 (λ) and u2 = u2 (λ) be the independent solutions of Equation (1) (or of Equa-
tion (2)). Show that the eigenvalue spectrum of L∗ in D∗ consists of the solution set of

< M B3 (u1 ), M B4 (u2 > − < M B4 (u1 , M B3 (u2 >= 0

Solution 14.2.8

(i) .....
(ii) needed to make the right hand side of Equation (6) vanish for all yr satisfying B1 (y1 ) = 0, B2 (y1 ) = 0.

To derive these conditions, we note that the right hand side of Equation (6) can be written in
T
matrix form as Y 2 M Y1 , where the matrix M is given by :
 
0 −p(a) 0 0
 
 
 p(a) 0 0 0 
 
M= 


 0 0 0 p(b) 
 
 
0 0 −p(b) 0

Let {B3 , B4 } be a spanning set for << B1 , B2 >>T . Then, for Y1 ∈ D, Y1T B1 = Y2T B2 = 0.

These imply the existence of complex numbers z1 , z2 such that Y1 = z1 B3 + z2 B4 Bq. It follows
that
T T T
Y 2 M Y1 = z1 Y 2 M B3 + z2 Y 2 M B4 (8)
T
Hence Y 2 M Y1 = 0 for Y1 ∈ D and Y2 ∈ D∗ , if and only if
T T
Y 2 M B3 = 0 and Y 2 M B4 = 0 (9)

These are the matrix form of the adjoint boundary conditions in D∗ .

79

A. Intissar and J-K. Intissar

Functional Analysis Spectral Theory of Linear Operators

Equivalently, the adjoint boundary conditions are specified by the fixed vectors

B1∗ = M B3 and B2∗ = M B4 (10)

The condition for an operator to be self-adjoint

(L, D) will be self-adjoint if its adjoint (L∗ , D∗ ) is such that L is symmetric and D∗ = D.

L given by equation (i) of (5) is symmetric, so that it will be self-adjoint , if and only if D∗ = D.

This will be so, if and only if

 ∗
 B1 = νB1 + ρB2
νσ − θρ 6= 0 (11)
 ∗
B2 = θB1 + σB2
or, equivalently, using result (10) and noting that M is nonsingular,

 B3 = νM −1 B1 + ρM −1 B2


νσ − θρ 6= 0 (12)
B4 = θM −1 B1 + σM −1 B2


Since B3 and B4 both belong to << B1 , B2 >>T

B1T B3 = B2T B3 = B1T B4 = B2T B4 = 0 (13)

Noting that M is antisymmetric, Equations (13) are consistent with Equation (12), if and only
if

B2T M −1 B1 = 0 (14)

which is the necessary and sufficient condition for L to be self-adjoint. The condition (14) is
the matrix form of the well known condition given in [Hellwig] :

p(a)(γ1 δ2 − γ2 δ1 ) = p(b)(α1 β2 − α2 β1 ) (15)

G. Hellwig, Differenlial Operators of Mathemaiical Physics, Addison Wesley Publishing CO.,

hc.,Fteading, Massachusetts, (1967).

(V) It is well known [[Hellwig], [Friedman]] that the eigenvalue spectrum of L in D is the
solution set of the following equation in λ

B1 (u1 )B2 (u2 ) − B2 (u1 )B1 (u2 ) = 0 (16)

where u1 = u1 (λ) and u2 = u2 (λ) are independent solutions of Equation (1) or of (2)).

B. Friedman, Principles and Techniques of Applied Mathematics, John Wiley and Sons, New
York, (1957).

Since L is symmetric, it follows that the eigenvalue spectrum of L∗ in D∗ consists of the solution
set of

80

A. Intissar and J-K. Intissar

Functional Analysis Spectral Theory of Linear Operators

M B3 (u1 )M B4 (u2 ) − M B4 (u1 )M B3 (u2 ) = 0 (17)

Without loss of generality, B3 and B4 can be chosen so that B3T M B4 = 0 (for if not, replace B4
0
by B4 = B4 − (B3T M B4 )(B3T B3 )−1 M −1 B3 . . Then both M B3 and M B4 are orthogonal to both
B3 and B3 so that

 M B3 = hB1 + kB2
hk 0 6= h0 k (18)
0 0
M B4 = h B1 + k B2


in which case Equation (17) can be written

(hB1 (u1 ) + kB2 (u2 ))(h0 B1 (u2 ) + k 0 B2 (u2 )) − (h0 B1 (u1 ) + k 0 B2 (u2 ))(hB1 (u2 ) + kB2 (u2 )) =

(hk 0 − h0 k)(B1 (u1 )B2 (u2 ) − B2 (u1 )B1 (u2 )) = 0.

Hence, Equations (16) and (17) have the same solution set and we have a direct proof that a
differential operator and its adjoint have the same eigenvalue spectrum [Naimark].

M.A. Naimark, Linear Difierential Operators, George G. Harrap Co., Ltd., London, (1966).

Exercise 14.2.9FFF

Part 1
Z x
Set S : C([0, 1], || . ||∞ ) −→ (R, | . |); f −→ (Sf )(x) = f (y)dy; f ∈ C([0, 1], || . ||∞ ), x ∈ [0, 1].
0

(i) Prove that S is continuous and calculate its spectrum σ(S).

(ii) Show that S is compact operator on C([0, 1], || . ||∞ ).

Part 2
Z 1−x
Set T : C([0, 1], || . ||∞ ) −→ (R, | . |); f −→ (T f )(x) = f (y)dy; f ∈ C([0, 1], || . ||∞ ), x ∈ [0, 1].
0

(iii) Prove that T is a linear bounded and compact operator on C([0, 1], || . ||∞ ).

(iv) Calculate the spectrum σ(T ) and the eigenvalues (σp (T )) of T .

Solution 14.2.9

Part 1

(i)

First we show that S is continuous. For f, g ∈ C[0; 1] we have

Z x
| Sf − Sg |=| S(f − g) |=| (f (y) − g(y))dy |
0

81

A. Intissar and J-K. Intissar

Functional Analysis Spectral Theory of Linear Operators

Z x
≤ | f (y) − g(y)) | dy
0
Z x
≤ supx∈[0,1] | f (y) − g(y)) | dy
0

≤|| f − g ||∞

Hence

|| S |≤ 1.
1
Next, we show that || S n || n −→ 0 as n −→ +∞. Observe that
x
x1
Z
| (Sf )(x) |≤ | f (y) | dy ≤|| f ||∞ x =|| f ||∞ , x ∈ [0, 1].
0 1!
By induction
x x
yn xn+1
Z Z
n+1 n
| (S f )(x) |≤ | S f (y) | dy ≤|| f ||∞ dy =|| f ||∞ , x ∈ [0, 1], n = 1, 2, ...
0 0 n! (n + 1)!
Thus,

1
|| (S n f ) ||∞ ≤|| f ||∞ n = 1, 2, ...
n!
So

1 1 1
|| S n || n ≤ ( ) n −→ 0 as n −→ +∞.
n!
1
Recall that the spectral radius of S is given by r(S) = lim || S n || n , asn −→ +∞.

This implies that r(S) = 0. Thus σ(S) = {0}.

(ii)

Let C ⊂ C[0; 1] is a bounded subset. Put || C ||= supf ∈C || f ||∞ .

Then S(C) is equicontinuous by the Fundamental Theorem of the Calculus :

0
∀f ∈ C, | (Sf ) (x) |≤|| f ||∞ ≤|| C ||.

S(C is bounded since :

∀f ∈ C, | (Sf )(x) |≤|| f ||∞ ≤|| C ||.

By Ascoli-Arzela theorem S(C) is relatively compact. Thus S is compact.

Part 2

(iii)

82

A. Intissar and J-K. Intissar

Functional Analysis Spectral Theory of Linear Operators

• Linearity of T : trivial.

• Boundedness of T :
Z 1−x Z 1−x Z 1−x
| (T f )(x) |=| f (y)dy |≤ | f (y) | dy ≤|| f ||∞ dy =|| f ||∞ (1 − x) ≤|| f ||∞ , where
0 0 0
|| f ||∞ = maxy∈[0,1] | f (y) |. Hence , || T ||≤ 1

• Compactness of T :

Let (fn )∞
n=1 be a bounded sequence in C([0, 1], || . ||∞ ) . Hence || fn ||∞ ≤ M, ∀ n ∈ N.

For some M ≥ 0. By Arzela-Ascoli theorem, it suffices to show that A := {T fn ; n ∈ N} is boun-

ded and equicontinuous. We have

|| T fn ||∞ ≤|| T || . || fn ||∞ ≤|| fn ||∞ ≤ M, ∀ n ∈ N.

Given any  > 0, without lost of generality, we can assume x < y, then
Z 1−x
| (T fn )(x) − (T fn )(y) |≤| f (y)dy |≤ M | x − y |.
1−y

| (T fn )(x) − (T fn )(y) |≤ ∀ n ∈ N provided | x − y |< .
M
Theorem 14.2.11 (Arzela-Ascoli Theorem)

Let (X; d) be a compact metric space and A ⊂ C(X). Then the following assertions are equiva-
lent :

1. A is relatively compact.

2. A is uniformly bounded and equicontinuous,

3. Any sequence (fn ) ⊂ A contains a uniformly convergent subsequence.

(iv)

First we see that λ = 0 is an eigenvalue. Assume that λ 6= 0 is an eigenvalue, i.e.,

Z 1−x
(T g)(x) = g(y)dy = λg(x), x ∈ [0, 1] for some 0 6= g ∈ C([0, 1], || . ||∞ ). This implies that
0

T g ∈ C 1 [0, 1] and λg 0 (x) = −g(1 − x), x ∈ [0, 1]. In particular if we put u = 1 − x we deduce
that λg 0 (u) = g(u),

Moreover, we have g(1) = 0. But g ∈ C 1 [0, 1] implies that g ∈ C 2 [0, 1].

By differentiating once more we get

g(x) x
λg 00 (x) = , x ∈ [0, 1] and g(1) = g 0 (0) = 0. Hence g(x) = Acos( ) with g(1) = 0. This gives
λ λ
1
that λk = π , k ∈ Z.
2 + kπ

83

A. Intissar and J-K. Intissar

Functional Analysis Spectral Theory of Linear Operators

Check if all these λ’s are eigenvalues. We calculate :

Z 1−x
t t π
(T gk )(x) = cos( )dt = λk [sin( )]1−x = λk sin[( + kπ)(1 − x)] = λk gk (x).
0 λk λk 0 2
Hence, λ = λ2l , l ∈ Z are the eigenvalues of T , i.e σp (T ) = {λ2l , l ∈ Z}.

[ σ(T ) is closed and as T is compact we have σ(T )/{0} ⊂ σp (T ). This yields

We know that
σ(T ) = {0} σp (T ).

Definition 14.2.12

Let T be a bounded operator on X where X is a Banach space. The spectral radius rσ (T )

of T is the radius of the smallest closed disk centered at the orgin and containing σ(T ).
rσ (T ) = supλ∈σ(T ) | λ |.

Formula for spectral radius :

1
It can be shown that rσ (T ) = lim || T n || n , n −→ +∞.

Exercise 14.2.10FF

Let H = L2 [0; 1] with the norm || . ||2 . We define the integral operator :
Z x
A : H −→ H by Af (x) = f (y)dy.
0

Show that A is bounded. Find || A ||.

Solution 14.2.10

Warning : The integral operator A is completely different from the same integral operator
defined on Banach space C([0, 1], || . ||∞ ).

First solution : Using Cauchy-Schwarz inequality, we get

Z 1 Z t
|| Af ||22 = | f (s)ds |2 dt
0 0
Z 1 Z tr
πs f (s)
= | .p cos ds |2 dt
0 0 2 cos πs2
Z 1 Z t Z t
πs | f (s) |2
≤ ( cos ds ds)dt
0 0 2 0 cos πs2
2 1 t πt | f (s) |2
Z Z
= ( sin ds)dt
π 0 0 2 cos πs 2
2 t 1 πt | f (s) |2
Z Z
= ( sin dt)ds
π 0 0 2 cos πs 2
2 t 1 | f (s) |2
Z Z
πt
= ( sin dt) ds
π 0 0 2 cos πs
2

84

A. Intissar and J-K. Intissar

Functional Analysis Spectral Theory of Linear Operators

t
| f (s) |2
Z
2
= ( )2 ds
π 0 cos πs
2
πs 2
Equality holds for f (s) = cos . Thus || A ||= π.
2
Exercise 14.2.11FFF

Part 1
Z 1 Z 1
Given N ∈ L2 [0; 1]×L2 [0; 1], i.e., | N (x, y) |2 dxdy < ∞, define the operator A : L2 [0; 1] −→ L2 [0; 1]
Z 1 0 0

by Af (x) = N (x, y)f (y)dy.

0

(i) Show that A is bounded..

(ii) Under what condition on N , the operator A is self-adjoint.

(iii) Show that A is compact.

Part 2

Let H = L2 [0; 1] with the norm || . ||2 . We define the integral operator :
Z x
A : H −→ H by Af (x) = f (y)dy.
0

(iv) Show that A is compact.

(v) Give the expression of the adjoint A∗ of A.

(vi) Calculate T = A∗ A.

(vii) Calculate the eigenvalues of T .

(viii) Compute || T || and || A || .

Solution 14.2.11

Part 1
Z 1 Z 1
(i) || Af ||2 = | N (x, y)f (y)dy |2 dx
0 0
Z 1 Z 1 Z 1
≤ ( | N (x, y) |2 dy)dx | f (y) |2 dy (Cauchy-Schwarz)
0 0 0
Z 1 Z 1
≤|| f ||2 . | N (x, y) |2 dydx.
0 0

Then :
s
Z 1 Z 1
|| A ||≤ || f ||2 . | N (x, y) |2 dydx < ∞.
0 0

85

A. Intissar and J-K. Intissar

Functional Analysis Spectral Theory of Linear Operators

Thus A is bounded.
Z 1 Z 1
(ii) < Af, g >= N (x, y)f (y)g(x)dx
0 0
Z 1 Z 1
= f (y) N (x, y)g(x)dxdy
0 0

< f, A∗ g >
Z 1
Thus we have A∗ g(x) = N (x, y)g(y)dy.
0

Hence, A is self-adjoint if N (x, y) = N (y, x).

∞
X
(iii) Let {ej (x)}∞
j=1 be an orthonormal basis in L2 [0, 1], then N (x, y) = < N (., y), ej > ej (x)
j=1
Z 1
where < N (., y), ej >= N (x, y)ej (x)dx := Nj (y) for almost all y.
0

Due to the Parseval identity, we have, for almost all y

Z 1 ∞
X
| N (x, y) |2 dx = | Nj (y) |2 .
0 j=1

and
Z 1 Z 1 ∞ Z
X 1
| Nj (y) |2 dxdy = | Nj (y) |2 dy.
0 0 j=1 0

Now, we define the following operator of rank k :

Z 1 k
X
Ak f (x) = Nk (x, y)f (y)dy where Nk (x, y) = Nj (y)ej (x).
0 j=1

By Cauchy-Schwarz inequality we obtain

Z 1 Z 1
2
|| (A − Ak )f || = | (N (x, y) − Nk (x, y))f (y)dy |2 dx
0 0
Z 1 Z 1 Z 1
≤( | | N (x, y) − Nk (x, y) |2 dxdy)( | f (y) |2 dy)
0 0 0
Z 1 Z 1
≤|| f ||2 ( | | N (x, y) − Nk (x, y) |2 dxdy)
0 0

Then
Z 1 Z 1
|| (A − Ak ) ||2 ≤ | N (x, y) − Nk (x, y) |2 dxdy
0 0
Z 1 Z 1
= [N (x, y) − Nk (x, y)][N (x, y) − Nk (x, y)]dxdy
0 0
Z 1Z 1
= [| N (x, y) |2 −N (x, y)Nk (x, y) − Nk (x, y)N (x, y)+ | Nk (x, y) |2 ]dxdy
0 0

86

A. Intissar and J-K. Intissar

Functional Analysis Spectral Theory of Linear Operators

Z 1 Z 1 Z 1 Z 1 Z 1 Z 1
2
= | N (x, y) | dxdy − N (x, y)Nk (x, y)dxdy − Nk (x, y)N (x, y)dxdy
0 0 0 0 0 0
Z 1 Z 1
+ | Nk (x, y) |2 dxdy
0 0
Z 1 Z 1 Z 1 Z 1 k
X Z 1 Z 1 k
X
= | N (x, y) |2 dxdy − N (x, y) Nj (y)ej (x)dxdy − N (x, y) Nj (y)ej (x)dxdy
0 0 0 0 j=1 0 0 j=1
k Z
X 1
+ | Nj (y) |2 dy
j=1 0
Z 1 Z 1 k Z
X 1
2
= | N (x, y) | dxdy − | Nj (y) |2 dy −→ 0 as k −→ +∞
0 0 j=1 0

All the Ak ’s are operators of finite rank and hence compact. Moreover, we have || A−Ak ||−→ 0
as k −→ +∞. But a uniform limit of a sequence of compact operators is compact. Hence, A is
compact.

Part 2
Z x Z 1
(iv) We write Af (x) = f (y)dy in the form Af (x) = N (x, y)f (y)dy
0 0

 1 if y ≤ x
where N (x, y) =
0 if y > x


We observe that N ∈ L2 [0; 1] × L2 [0; 1] and N (x, y) 6= N (y, x) then by applying (iii), we deduce
that A is compact but it is not self-adjoint.
Z 1
(v) < Af, g >= Af (x)g(x)dx
0
Z 1 Z x
= f (s)dsg(x)dx
0 0
Z x Z x Z 1 Z x
=[ f (s)ds g(s)ds]10 − f (x)( g(s)ds)dx
0 0 0 0
Z 1 Z 1 Z 1 Z 1 Z 1
= f (s)ds g(s)ds + f (x)[ g(s)ds − g(s)ds]dx
0 0 0 x 0
Z 1 Z 1
= f (x)( g(s)ds)dx
0 x
Z 1
We have that A∗ is given by A∗ f (x) = f (y)dy.
x
Z 1 Z 1 Z y
(vi) Hence from (ii) we get A∗ Af (x) = Af (y)dy = [ f (s)ds]dy = T f (x).
x x 0

(vii) T f is twice differentiable for all f ∈ H, so any eigenvector of T λf (x) = T f (x) must satisfy
the differential equation :

d2 d2
λ f (x) = T f (x)
dx2 dx2

87

A. Intissar and J-K. Intissar

Functional Analysis Spectral Theory of Linear Operators

Z 1
d2
= 2[ Af (y)dy]
dx x
Z 1Z y
d2
= 2 f (s)dsdy
dx x 0
Z x
d
= (− f (s)ds)
dx 0

= −f (x), λ ∈ C/{0}.

By a theorem from differential equations, we know that there are only two linearly independent
solutions to the above differential equation over C, namely, eiωx and e−iωx , where ω 2 = λ such
that λ, ω 6= 0. Hence the general solution to the above equation is :

f (x) = αeiωx + βe−iωx , α, β ∈ C.

When we apply T , we get :

Z 1 Z y
−iωx
T (αe iωx
+ βe )= (αeiωs + βe−iωs )dsdy
x 0
1
αeiωs βe−iωs y
Z
= [ − ] dy
x iω iω 0
1
αeiωy βe−iωy
Z
α β
= [( − ) − ( − ]dy
x iω iω iω iω
αeiωy βe−iωy α β
= [( − ) − ( − ]y]1x
−ω 2 −ω 2 iω iω
αeiω βe−iω α β αeiωx βe−iωx α β
= [( 2
− 2
) − ( − ) - [( 2
− 2
) − ( − )x]
−ω −ω iω iω −ω −ω iω iω
1 iωx −iωx α β αeiω βe−iω α β
= 2
(αe + βe ) + ( − )x − ( 2
− 2
)−( − )
ω iω iω ω ω iω iω
Now since αeiωx + βe−iωx is an eigenvector, we must have that

α β αeiω βe−iω α β
( − )( 2 − ) − ( − ) = 0, ∀ x ∈ [0, 1]
iω iω ω ω2 iω iω
which implies α = β. Hence αeiωx + βe−iωx = 2αcosωx.

Moreover, we must have that T f (1) = 0 for any eigenvector f by, so 2αcosω = 0. Since α 6= 0,
we must have that

(2n + 1)π
ω= , n ∈ Z.
2
Hence the eigenvalues are of the form

1 4
λn = = , n ∈ Z.
ωn2 (2n + 1)2 π 2
(viii) Since A is compact and T = A∗ A is self adjoint, we have that T is a compact, normal
operator. Hence, its spectrum is countable with 0 as the only possible cluster point and || T ||
is equal to the spectral radius r(T ) of T where r(T ) is the radius of the smallest closed disk

88

A. Intissar and J-K. Intissar

Functional Analysis Spectral Theory of Linear Operators

centered at the origin and containing σ(T ) i.e. r(T ) = max{λ ∈ σ(T )}.
4 4 p 2
As max{λn ; n ∈ Z} = 2
(take n = 0, −1). We conclude that || T ||= 2 and || A ||= || T || = .
π π π
F
Exercise 14.2.12

We consider the Sturm-Liouville equation

d d
[p(x) y] + (q(x) + λr(x))y = 0 (E)
dx dx
Where p(x), q(x), r(x) are given functions defined on [a, b], p(x) 6= 0, y = y(x) is a function to
be determined, λ is a constant to be determined as well.

Let the functions a(x), b(x) and c(x) defined on the interval [a; b] are such that a(x) 6= 0 for all
b(x)
x ∈ [a; b] and that the function is integrable on [a; b], and if we define :
a(x)
R x b(t)
c(x) p(x)
p(x) = e a a(t) dt , q(x) = p(x) and r(x) = .
a(x) a(x)
Show that the Sturm-Liouville equation (E) is equivalent to the following ordinary linear dif-
ferential equation of order 2 :

d2 y dy
a(x) 2
+ b(x) + (c(x) + λ)y = 0 , x ∈ [a, b] (EE)
dx dx
Solution 14.2.12

d d d2 dp dy d2 y b(x) Rax b(t)

dt dy d2 y b(x) dy
[p(x) y] = p(x) 2 y + = p(x) 2 + [e a(t) ] = p(x) 2 + p(x) .
dx dx dx dx dx dx a(x) dx dx a(x) dx
Substituting this, as well as the expressions for p(x) and r(x) in the Sturm-Liouville equation,
we obtain

d2 y b(x) dy c(x)p(x) p(x)

p(x) 2
+ p(x) +( +λ )y = 0.
dx a(x) dx a(x) a(x)
=⇒

d2 y dy
a(x) + b(x) + (c(x) + λ)y = 0 , x ∈ [a, b].
dx2 dx
a(x)
We obtain this last equation by multiplying both sides of the first equation by . Because
p(x)
p(x) is the exponential of a real number, then p(x) > 0 and so the division by p(x) for all
x ∈ [a, b] is well defined. In particular if y = y(x) is a solution of the Sturm-Liouville equation
(E), then there will also be a solution to equation (EE).

The converse is also true, ie if y is a solution of equation(EE), then y is a solution of the

Sturm-Liouville equation (E) with the functions p(x), q(x) and r(x) as defined above..

Definition 14.2. 13

A problem of Sturm-Liouville is an equation of Sturm-Liouville

89

A. Intissar and J-K. Intissar

Functional Analysis Spectral Theory of Linear Operators

d d
[p(x) y] + (q(x) + λr(x))y = 0 (E)
dx dx
where p > 0, r ≥ 0, and p, q, r are continuous functions on interval [a, b], with in addition the
boundary conditions at the ends of the interval [a ; b] of the form :

a1 y(a) + a2 p(a)y 0 (a) = 0, b1 y(b) + b2 p(b)y 0 (b) = 0, (BC)

where a21 + a22 6= 0 and b21 + b22 6= 0.

The problem of finding a complex number µ if any, such that the Boundary Value Problem
((E)-((BC)

with ?λ = µ, has a non-trivial solution is called a Sturm-Liouville Eigen Value Problem (SL-
EVP). Such a value µ is called an eigenvalue and the corresponding non-trivial solutions y(.; µ)
are called eigenfunctions.

Further,

(α) An (SL-EVP) is called a regular (SL-EVP) if p > 0 and r > 0 on[a, b].

(β) An (SL-EVP) is called a singular (SL-EVP) if

(i) p > 0 on (a, b) and p(a) = 0 = p(b),

and

(ii) r ≥ 0 on [a, b].

(γ) If p(a) = p(b), p > 0 and r > 0 on [a, b], p, q, r are continuous functions on [a, b], then solving
Sturm-Liouville equation (E) coupled with boundary conditions

 y(a) = y(b),
(Periodic Conditions)
 0
y (a) = y 0 (b)

is called a periodic (SL-EVP).

Remark 14.2. 14

We are not going to discuss in this sheet general Sturm-Liouville problem as it was announced
in Sheet 1 around the following points :

• Defining the minimal domain Dmin and the maximal domain Dmax of the operator `.

• What conditions are required to ensure that Dmin = Dmax .

• What Left-Definite Theory of bounded-below, self-adjoint operators in a Hilbert space.

• What boundary conditions are required to ensure that the unbounded operator (`, L2 [(a, b), w])
is self adjoint.

90

A. Intissar and J-K. Intissar

Functional Analysis Spectral Theory of Linear Operators

This discussion is reported on Sheet 5. However, in next exercise, we discuss completely an

example of singular Sturm-Liouville which is attached to Reggeon field theory in Bargmann
representation.

Exercise 14.2.12FFFF
Z
2
Let B = {ϕ : C −→ C entire ; | ϕ(z) |2 e−|z| dxdy < ∞, ϕ(0) = 0} (called Bargmann space).
C
0 3
Given (λ , µ, λ) ∈ R , we consider on B the following eigenvalue problem :

 Hϕ = σϕ, on D(H)
(?)
σ∈C


Where

• D(H) = {ϕ ∈ B; Hϕ ∈ B}
00 0
• Hϕ(z) = (λ0 z 2 + iλz)ϕ (z) + (iλz 2 + µz)ϕ (z); i2 = −1.
 
ϕ(z)
0
(1) Put ψ(z) = ϕ (z) in (?) and φ(z) =  
ψ(z)
Show that
 
0 1
0  
φ (z) = 
 σ iλ + µ  φ(z)

− 0
z(λ0 z + iλ) λ z + iλ
ϕ(reiθ )
 

(2) Put z = reiθ ; r ∈ [0, +∞[, θ ∈ [0, 2π] and W (r) =  

iθ
ψ(re )
Show that

W 0 (r) = e−iθ M(reiθ )W (r) (??)

 
0 1
 
where M(reiθ ) =  0

 σ(λ , µ, λ) iλ + µ 
−
reiθ (λ0 reiθ + iλ) λ0 reiθ + iλ

(3) In C2 we denote the norm of W by || W || and the norm of matrix M by ||| M |||.

λ 0
Let R > ρ0 = ; λ 6= 0. Show that the norm ||| M ||| is bounded outside the disk of radius R
λ0
and ∀ r ∈ [R, +∞[, we have || φ(z) ||≤ supθ∈[0,2π] || W (Reiθ ) || e|||M|||(|z|−R) . In particular, we
get | ϕ(z) |≤ supθ∈[0,2π] || W (Reiθ ) || e|||M|||(|z|−R)

91

A. Intissar and J-K. Intissar

Functional Analysis Spectral Theory of Linear Operators

(4) Let u(y) = ϕ(−iy) show that

Hu(y) = −y(λ − λ0 y)u” (y) + y(λy + µy)u0 (y) (???)

(5) Let ρ0 > 0, y ∈ [0, ρ0 ] and L2 ((0, ρ0 ), r(y)dy) be the Hilbert space of square integrable
Z ρ0
0
functions on [0, ρ ] with inner product < u, v >= u(y)v(y)r(y)dy with r(y) called weight
0
function.

Find the fonction r(y) such that H defined by (???) is symmetric on L2 ((0, ρ0 ), r(y)dy).

(6) Show that on L2 ((0, ρ0 ), r(y)dy), the operator H takes the following form

1 d
Hu(y) = [p(y)u0 (y)] where p(y) = λ0 y(ρ0 − y)r(y) (????)
r(y) dy
(7) Show if ρ0 (ρ + ρ0 ) ≥ 1 then the eigenvalues of H acting on L2 ((0, ρ0 ), r(y)dy) are real and
simple.

(8) Open question : On L2 ((0, ρ0 ), r(y)dy) H is equivalent to self-adjoint operator or to normal

operator. 

Solution 14.2.12

(8)
(1) We consider the equation
00 0
(λ0 z 2 + iλz)ϕ (z) + (iλz 2 + µz)ϕ (z) = σϕ(z) (?)
0 00
In our differential equation (?), if we put ψ(z) = ϕ (z) then ψ 0 (z) = ϕ (z) and we get

0 σ(λ0 , µ, λ) iλ + µ
ψ (z) = ϕ(z) − 0 ψ(z)
z(λ0 z + iλ) λ z + iλ
 
  0 1
ϕ(z)
0  
Now let φ(z) =   then φ (z) = 
 σ(λ0 , µ, λ)
 φ(z)
iλ + µ 
ψ(z) − 0
z(λ0 z + iλ) λ z + iλ
ϕ(reiθ )
 

(2) If we put z = reiθ ; r ∈ [0, +∞[, θ ∈ [0, 2π] and W (r) =  , we deduce that
iθ
ψ(re )
 
0 1
W 0 (r) = e−iθ M(reiθ )W (r) where M(reiθ ) = 
 
0

 σ(λ , µ, λ) iλ + µ 
−
reiθ (λ0 reiθ + iλ) λ0 reiθ + iλ

(3) In C2 we denote the norm of W by || W || and the norm of matrix M by ||| M |||.

As || M(reiθ )W (reiθ ) || ≤ ||| M ||| || W ||, then we deduce that

92

A. Intissar and J-K. Intissar

Functional Analysis Spectral Theory of Linear Operators

0
|| M (r) ≤||| M ||| || W (r) ||

Let R > ρ0 then the norm ||| M ||| is bounded outside the disk of radius R and by applying the
Gronwal Lemma, we deduce the following inequality

∀ r ∈ [R, +∞[, || φ(z) ||≤ supθ∈[0,2π] || W (Reiθ ) || e|||M|||(|z|−R)

In particular, we get | ϕ(z) |≤ supθ∈[0,2π] || W (Reiθ ) || e|||M|||(|z|−R)

Therefore ϕ belongs to Bargmann space B0

(4) Let u(y) = ϕ(−iy) then u0 (y) = −iϕ0 (−iy) and u00 (y) = −ϕ00 (−iy) this implies that

Hu(y) = −y(λ − λ0 y)u” (y) + y(λy + µy)u0 (y) (???)

(5) Let ρ0 = > 0, y ∈ [0, ρ0 ] and L2 ((0, ρ0 ), r(y)dy) be the Hilbert space of square integrable
λ
λ0
Z ρ0
functions on [0, ρ0 ] with inner product < u, v >= u(y)v(y)r(y)dy with r(y) called weight
0
function.

We will construct the fonction r(y) such that H is symmetric on L2 ((0, ρ0 ), r(y)dy).

For ϕ ∈ B0 , consider u(y) = ϕ(−iy) and the eigenvalue problem associated to Hλ0 ,µ,λ :

−y(λ − λ0 y)u” (y) + y(λy + µy)u0 (y) = σu(y)

multiplying the differential equation by r(y)u(y) and integrating it on [0, ρ0 ] to get

Z ρ0 Z ρ0
[−y(λ − λ0 y)u” (y) + y(λy + µy)u0 (y)]u(y)r(y)dy = σ | u(y) |2 r(y)dy
0 0

Integrating the first term of integral equation by parts to get

0
Z ρ0
[s(y)u(y)u (y)(λ y − 0 0
λ)]ρ0 − s(y)(λ0 y − λ) | u0 (y) |2 dy
0
Z ρ0 Z ρ0
∂
+ ((λ0 y − λ)s(y))] | u(y) |2 dy = σ
[(λy + µ)s(y) − | u(y) |2 r(y)dy
0 ∂y 0
∂ 0
where s(y) = yr(y). We choose s(y) such that ((λ y − λ)s(y)) = (λy + µ)s(y)
∂y
0 0 0 s(y)
Then s(y) = ceρ y | y − ρ0 |ρ (ρ+ρ )−1 and r(y) = , c is a constant.
y
Now, if ρ0 (ρ + ρ0 ) ≥ 1 then the function s(y) is bounded on [0, ρ0 ] this implies that
0
[s(y)u(y)u0 (y)(λ0 y − λ)]ρ0 = 0

Consequently
Z ρ0 Z ρ0
0 0 2
− s(y)(λ y − λ) | u (y) | dy = σ | u(y) |2 r(y)dy
0 0

93

A. Intissar and J-K. Intissar

Functional Analysis Spectral Theory of Linear Operators

i.e H is symmetric on L2 ((0, ρ0 ), r(y)dy). 

(6) On L2 ((0, ρ0 ), r(y)dy), it is trivial that the operator H takes the following form

1 d
Hu(y) = [p(y)u0 (y)] where p(y) = λ0 y(ρ0 − y)r(y) (????)
r(y) dy
(7) If ρ0 (ρ + ρ0 ) ≥ 1 we have (see (5))
Z ρ0 Z ρ0
− s(y)(λ0 y − λ) | u0 (y) |2 dy = σ | u(y) |2 r(y)dy
0 0

then the the eigenvalues of H are real.

Now, we show that if ρ0 (ρ + ρ0 ) ≥ 1 then the eigenvalues of Hλ0 ,µ,λ acting on L2 ((0, ρ0 ), r(y)dy)
are simple.

In fact :

v ∂
Let u, v ∈ B1 where B1 = {u ∈ L2 ((0, ρ0 ), r(y)dy); u(0) = 0} then vHu = (p(y)u0 (y)),
r(y) ∂y
u ∂
uHv = (p(y)v 0 (y))
r(y) ∂y
and

∂
r(y)(vHu − uHv = [p(y)(v(y)u0 (y) − u(y)v 0 (y)]. (•1 )
∂y
If Hu(y) = σu(y); u 6= 0 and Hv(y) = σv(y); v 6= 0 then

∂
[p(y)(v(y)u0 (y) − u(y)v 0 (y)] = 0, in particular p(y)(v(y)u0 (y) − u(y)v 0 (y) is a constant.
∂y
As p(0)((v(0)u0 (0) − u(0)v 0 (0)) = 0 then v(y)u0 (y) − u(y)v 0 (y) = 0 ∀ y ∈ [0, ρ0 ] and v is pro-
portional to u i.e. the geometric multiplicity of σ is one.

Now, let Ker(H − σI) = {u ∈ B1 ; Hu = σu; u 6= 0}, showing that there are no Jordan blocks
corresponding to σ i.e ;

Ker(H − σI) = Ker(H − σI)2 , ∀ σ ∈ σ(H) where σ(H) denotes the spectrum of H.

Let u ∈ Ker(H − σI) and v ∈ Ker(H − σI)2 such that

1 ∂
(p(y)u0 (y))u(y) = σu(y), y ∈ [0, ρ0 ]
r(y) ∂y
and

1 ∂
(p(y)v 0 (y))v(y) = σv(y) + u(y), y ∈ [0, ρ0 ] (•2 )
r(y) ∂y
Multiplying the differential equation (•2 ) by r(y)v(y) and by r(y)v(y), integrating by parts the
obtained equations on [0, ρ0 ] to get

94

A. Intissar and J-K. Intissar

Functional Analysis Spectral Theory of Linear Operators

Z ρ0 Z ρ0
0 0
p(y)u (y)v (y)dy = σ r(y)u(y)v(y)dy (•3 )
0 0

and
Z ρ0 Z ρ0 Z ρ0
p(y)v 0 (y)u0 (y)dy = σ r(y)v(y)u(y)dy + r(y) | u(y) |2 dy (•4 )
0 0 0

Now (•3 ) - (•4 ) gives

Z ρ0 Z ρ0 Z ρ0
0 0 0 0
p(y)[u (y)v (y) − v (y)u (y)]dy = σ r(y)[u(y)v(y) − v(y)u(y)]dy − r(y) | u(y) |2 dy
0 0 0

In particular we have
Z ρ0 Z ρ0 Z ρ0
2i p(y)=m(u0 (y)v 0 (y))dy = 2iσ r(y)=m(u(y)v(y)) − v(ydy − r(y) | u(y) |2 dy
0 0 0
Z ρ0
As p is real function and σ ∈ R then r(y) | u(y) |2 dy = 0 which is contradictory with u is
0
eigenfunction associated to σ. 

(8) Open question : For the authors of this exercise, all reader’s contribution on this topic
is welcome !

Light-Cone

95
View publication stats
A. Intissar and J-K. Intissar