Functional Analysis Basics

c 2007 Ulrich G. Schuster

Vector Space Concepts

incomplete

linear spaces

complete

normed spaces
Banach space
inner product spaces

Hilbert space

metric spaces

Scalar Field. For all linear (vector) spaces

in the following, the scalar field will be either the field of real numbers, F = R, or the
complex field, F = C.

If a normed space U contains a sequence {en } with the property that for
every u U there is a unique sequence of
scalars {n } such that ku(1 e1 +2 e2 +
+N eN k 0 as N , then {en } is
called a Schauder basis for U. A Schauder
basis is different from a Hamel basis in
that a countably infinite number of basis vectors and scalar coefficients may be
needed to uniquely represent a given vector.

Normed Space [1, 2, 2]. A norm kk on

a linear space (U, F) is a mapping kk : U
[0, ) that satisfies, for all u, v U, F,
1. kuk = 0 u = 0.
2. kuk = || kuk.
3. Triangle inequality: ku + vk kuk +
kvk.
A norm defines a metric d(u, v) := ku vk Convergence [1, 4.8]. Let {un } be a
on U. A normed (linear) space (U, kk) is a sequence of vectors in a normed space U.
linear space U with a norm kk defined on The sequence {un } is said to be strongly
it.
convergent, or convergent in norm, if
The norm is a continuous mapping of U
there is a u U, called the strong limit
into R+ .
of {un }, such that limn kun uk = 0.
A norm kk on a linear space U is said
Strong convergence is written un u
to be equivalent to a norm kk0 on U if
and often referred to simply as converthere are positive numbers a and b such
gence.
that akuk0 kuk bkuk0 for all u U. The sequence {un } is said to be weakly
Equivalent norms define the same topolconvergent if there is a u U,
ogy on U.
called the weak limit of {un }, such
The metric d induced by a norm is transthat limn f (un ) = f (u) for every
lation invariant, i.e., it satisfies
bounded linear functional f on U, i.e.,
0
d(u + x, v + x) = d(u, v),
for every f in the dual space U . Weak
w
d(u + v) = || d(u, v)
convergence is written un
u.
for all u, v, x U and F.
The weak limit u is unique.
Rieszs Lemma: Let Y and Z be linear
Every subsequence of {un } converges
subspaces of a normed space U and let Y
weakly to u.
be a closed proper subset of U. Then, for
The sequence {kun k} is bounded.
every (0, 1), there is a z Z such
that kz yk for kzk = 1 and for Strong convergence implies weak convergence to the same limit.
all y Y.
A subset T of a normed space U is said If dim U < , then weak convergence
implies strong convergence.
to be total in U if span T is dense in U.
Let S be a linear subspace of a normed The (infinite) series u1 + u2 + . . . is said
to converge (strongly) if the sequence of
space U. If S is open as a subset in U,
partial sums sn := u1 + u2 + + un
then S = U.
converges, i.e., if sn s for some s U.
Basis and Dimension [2, 1].
The above series is said to be absolutely
Let V be a linear space. A linearly inconvergent if the infinite series ku1 k +
dependent subset S V that spans V is
ku2 k + . . . converges.
called a Hamel basis for V.
A series is said to be unconditionally convergent if (i) it is convergent for each pos Every linear space has a Hamel basis,
sible rearrangement of terms, and (ii) if
so that every nonzero v V has a
each rearrangement converges to the same
unique representation as a linear comlimit.
bination of finitely many elements of S
with nonzero scalar coefficients.
Banach Space [1, 2].
A Banach
If B1 and B2 are Hamel basis for a linspace (B, kk) is a complete normed space,
ear space V, then they have the same
complete in the metric induced by its
cardinality.
norm kk.
The dimension dim V of a linear space V
is defined as the cardinality of any Hamel A linear subspace S of a Banach space B
is a Banach space, i.e., it is complete, iff S
basis of V.
is closed in B.
If dim V is finite, V is called a finite For a series on a Banach space, absolute
dimensional linear space.
convergence implies strong convergence
A linear space V is finite dimensional
and unconditional convergence.
iff there is a positive integer N such
Let (U, kk) be a normed space. Then
that V contains a linearly indepenthere is a Banach space B and an isomedent set of N vectors whereas any set
try f from B onto a linear subspace S B
of N + 1 vectors of V is linearly depenthat is dense in B. The space B is
dent.
unique except for isometries. Thus, every
If V1 and V2 are linear spaces over the
normed space can be completed.
same scalar field, then they are isomorphic iff dim V1 = dim V2 .

SVN Revision: 2109, December 11, 2007

Finite-Dimensional Normed Spaces.
complete, then A is complete.
Every finite-dimensional linear sub- If A B, then B A .
space S of a normed space U is complete; A (A ) .
in particular, every finite-dimensional If g A A , then g = 0.
normed space is complete.
If A G, then A = ((A ) ) .
Every finite-dimensional linear subspace {0} = G and G = {0}.
of a normed space U is closed in U and If A is a dense subset of G, then A =
separable.
{0}.
On a finite-dimensional linear space, all If {An } is a sequence of sub

norms are equivalent.
spaces, then span{An }
= n A
n,
In a finite-dimensional normed space U,
any subset S U is compact iff S is
}.
and (n An ) = span{A
n
closed and bounded.
An orthonormal set O in an inner product
Inner Product Space [1, 3]. Let (G, F) space G that is total in G is called a total
be a linear space. An inner product is a orthonormal set, or sometimes a maximal or
mapping h, i : G G F that satisfies the complete orthonormal set.
following properties for all x, y, z V and Let O G be a subset of an inner product
space G. Then, if O is total in G, there
scalars F:
does not exist a nonzero vector g G
1. hx + y, zi = hx, zi + hy, zi.
that is orthogonal to every element of O.
2. hx, yi = hx, yi.
If G is complete, i.e., a Hilbert space, the
3. hx, yi = hy, xi .
above condition is sufficient for O to be
4. hx, xi 0 with equality iff x = 0.
total in G.
A linear space G on which an inner product h, i is defined is called an inner product Hilbert Space [1, 3]. A complete inner
space (G, h, i).
product space (H, h, i) is called a Hilbert
An
p inner product defines a norm kxk := space. Thus, a Hilbert space is a Banach
hx, xi
p and a metric d(x, y) := ky space on which an inner product is defined.
xk = hy x, y xi on G. Hence, in- For any inner product space G there exists a Hilbert space H and an isomorner product spaces are normed spaces.
phism from G onto a dense linear sub The inner product is called sesquilinspace D H. The space H is unique
ear, because it is linear in the first
except for isomorphisms. Thus, every
term and conjugate linear in the second
inner product space can be completed.
term: hx, yi = hx, yi.
The inner product satisfies the Schwarz Let {hn } be a sequence in a Hilbert
w
inequality: |hx, yi| kxkkyk.
space H. Then, hn
h iff hhn , zi
The induced norm satisfies the triangle inhh, zi for all z H.
equality: kx + yk kxk + kyk with equal- In every Hilbert space H =
6 {0}, there
ity iff y = cx for some positive scalar c.
exists a total orthonormal set.
The induced norm satisfies the parallel- An orthonormal set O in a Hilbert
ogram equality: kx + yk2 + kx yk2 =
space H is total in H iff for all h H the
2(kxk2 + kyk2 ).
Parseval relation
holds:
X
2
Continuity: if in an inner product space G
|hh, ei| = khk2 .
xn x and yn y, then hxn , yn i
eO
hx, yi, where {xn }, x, {yn }, y G.
If hx1 , yi = hx2 , yi for all y in an inner A total orthonormal sequence, i.e., a
countable total orthonormal set, in a
product space, then x1 = x2 .
Hilbert space H is called an orthonormal
Two inner product spaces G and V are called
basis
for H.
unitarily equivalent if there is an isomor
If
a
Hilbert
space H is separable, every
phism U : G V of G onto V that preserves
total
orthonormal
set is countable, i.e., evinner products, i.e., hUu1 , Uu2 i = hu1 , u2 i
ery total orthonormal set is an orthonorfor all u1 , u2 G. The mapping U is called
mal basis. Conversely, if H contains an
a unitary operator.
orthonormal sequence that is total in H,
Orthogonality [2, 1]. An element x of an
then H is separable. Thus, there exists
inner product space G is said to be orthogan orthonormal basis for H iff H is sepaonal to an element y G, denoted x y,
rable.
if hx, yi = 0. Similarly, for A, B G, All total orthonormal sets in a given
x A means that x a for all a A,
Hilbert space have the same cardinality,
and A B means that a b for all a A
called the Hilbert dimension of H.
both over
and all b B.
Two Hilbert spaces H and H,
the
same
scalar
field,
are
isomorphic
iff
An orthogonal set O in an inner product
they have the same Hilbert dimension.
space G is a subset O G whose elements
are pairwise orthogonal. An orthonormal Let Y be any closed linear subspace of
a Hilbert space H. Then, H = Y Z,
set is an orthogonal set whose elements
where Z = Y is the orthogonal complehave unit norm. A countable orthogonal
ment of Y. Each h H can be uniquely
(orthonormal) set is called an orthogonal
represented as h = y + z, where y Y
(orthonormal) sequence.
and z Z = Y , and khk = kyk + kzk.
An orthogonal set is linearly independent.
Let {e } be an orthonormal set in an Let S H be a linear subspace of H;
inner product space G, and let g be any
then, (S ) = S. If S is closed,
point in G. Then hg, e i is nonzero for at
then (S ) = S.
most a countable number of vectors e .
For any nonempty susbspace S of a
Let G be an inner product space and C a
Hilbert space H, span S is dense in H
nonempty convex subset of G that is comiff S = {0}. If S is closed and S =
plete in the metric induced by the inner
{0}, then S = H.
product. Then, for every g G there ex- Let Y and Z be two subspaces of H. The
ists a unique c0 C such that inf cC kg canonical correlation (Y, Z) between these
ck = kg c0 k. If C is a complete linear two subspaces is defined as
subspace of G, then (g c0 ) C.

(Y, Z) := sup |hy, zi| : y Y,
Bessel inequality: Let {en } be an or
thonormal sequence in an inner product
z Z, kyk = kzk = 1
space G. Then, for every g G,
and the angle (Y, Z) between these sub
X
2
spaces as (Y, Z) = cos (Y, Z).
|hg, en i| kgk2 .
Let y Y, z Z. Then, the following
n=1
conditions are equivalent:
Orthogonal Complement [2, 1, 3].
(Y, Z) < 1, i.e., (Y, Z) > 0.
Let A and B be nonempty subsets in an
inf {ky zk : kyk = kzk = 1} > 0.
inner product space G. The set A :=
There is a constant c such that kyk
{g G : g A} is called the orthogonal
cky + zk for all y, z.
complement of A in G.

The direct sum Y Z is a closed sub The orthogonal complement A of A in G

space of H.
is a closed linear subspace of G. If G is

Fourier Series [2]. Riesz-Fischer Theorem: Let {en } be an orthonormal sequence

in a Hilbert space H, and let {n } be a
sequence of scalars. Then, the series

X
h=
n en
n=1

converges in norm iff n=1 |n | < . In

this case, the coefficients n are called the
Fourier coefficients of h, and they are given
as n = hh, en i. Conversely, the above series always converges to h if the n are the
Fourier coefficients of any h H.
The above series is convergent iff it converges unconditionally.
Let {en } be an orthonormal set in a Hilbert
space H, then the following statements are
equivalent:
The set {en } is an orthonormal basis
for H.
For any h H, the Fourier series
exP
pansion of h is given as h = n n en ,
where n = hh, en i.
Parseval equality: For any x, y H,
X
hx, yi =
hx, en ihy, en i .

An element b B is called invertible if b

has an inverse in B. The invertible elements of B form a group with respect to
vector multiplication.
Let S B denote the set of all invertible
elements of B. If b B and kbk < 1,
then,
1 + b S, P

(1 + b)1 = n=0 (1)n bn ,

k(1 + b)1 1 + bk kbk2 /(1 kbk).
The set S is open, and the mapping b b1 is a homeomorphism
of S onto S.
The spectrum S(b) of an element b B
is defined as the set of all complex numbers such that b 1 is not invertible.
Let f be a bounded linear functional on B.
Then, for any fixed b B, the function g() := f ((b 1)1 ),
/ S(b), is
holomorphic in the complement of S(b),
and g() 0 as .
For every b B, the spectrum S(b) is
compact and not empty.
If each nonzero element of B is invertible, then the complex Banach algebra B
is isometrically isomorphic to the complex field C. This also implies that B is
commutative.
For any b B, the spectral radius rb of b is defined as rb :=
sup {|| : S(b)}; it can be computed
as r= limn kbn k1/n .
A complex-valued homomorphism f on a
Banach algebra B is a linear functional that
preserve vector multiplication, i.e., a functional f for which f (x + y) = f (x) +
f (y) and f (xy) = f (x)f (y) for all x, y
B and , F. Furthermore, f is not
identical to 0. Let M denote the set of
all complex-valued homomorphisms f of B.
Then,
S(b) iff f (b) = for some f M.
The vector b is invertible in B iff f (b) 6= 0
for every f M.
f (b) S(b) for every b B and f M.
|f (b)| rb kbk for every b B
and f M.

Hardy Space [3].

Let D :=
{z C : |z| < 1} be the open disk in the
complex plane. For 0 < p < , the space
n
Hp := f : f analytic in D,
Z
o


f (rei )p d <
sup
0r<1

with norm
Z
kf kp := sup



f (rei )p d

1/p

0r<1

is a Banach space, called the Hardy space.

Reproducing Kernel Hilbert Spaces.
to write

Linear Operators and Linear Functionals

Linear Operator [1, 2]. A linear operadisjoint subspaces of X such that X =

tor T is a mapping of a linear space V into
R(P) + N (P) = R(P) N (P), i.e., R(P)
a linear space Z such that
and N (P) are algebraic complements of
one another.
1. The domain D(T) is a linear space V,
and the range R(T) lies in a linear If P is a projection, so is IP, and R(P) =
space Z over the same scalar field F.
N (I P) and N (P) = R(I P).
2. For all v, u V and scalars ,
Let S X be a subspace of X . Then
there exists a projection P : X X such
T(v + u) = Tv + Tu,
that R(P) = S.
T(v) = Tv.
Given two disjoint subspaces V and U
The null space N (T) of T is the set of
with X = U V, there is a unique projecall v D(T) such that Tv = 0. The null
tion P such that R(P) = U and N (P) =
space is a linear space.
V.
The range space R(T) of a linear operator
Finite-Dimensional Spaces [1, 2.9,
is a linear space.
7.1].
Let X and Y be finite Two linear operators T and S are said to
n
dimensional
linear
spaces over the same
be equal if they have the same domain
field
F,
with
dim
X
= N, dim Y = K.
For any h H,
and if Tv = Sv for all v D(T) = D(S).
X
Let E := {e1 , . . . , eN } be a basis for X , and
2
2

dim
D(T)
=
N
<

=
dim
R(T)

khk =
|hh, en i| .
let B := {b1 , . . . , bK } be a basis for Y.
N.
n
The dimensions of the null space N (T),, Any linear operator T : X Y is
Let M be any linear subspace of H that
uniquely determined by the K images
the range space R(T) and the space X itcontains {en }; then M is dense in H.
of the N basis vectors yk = Ten .
self are related as dim N (T)+dim R(T) =
Any linear operator T on a finiteBanach Algebra [4, 5]. Strictly speakdim X .
dimensional linear space can be repreing, a Banach Algebra is an algebra B over
Let X1 , X2 , Y1 , Y2 be linear spaces over
sented by a matrix T with [T]k,n = tk,n ,
a scalar field F, where B is also a Banach
the same scalar field so that X1 and X2
where T depends on the bases E and B.
space under a norm kk that satisfies the
are isomorphic and Y1 and Y2 are isomorHence, the image of any vector x X
multiplicative inequality kxyk kxkkyk
phic. The linear operators T1 : X1 Y1
can be obtained as
for all x, y B.
and T2 : X2 Y2 are said to be isomorK X
N
In the following, though, an associative unit
X
phically equivalent if there exists isomorcomplex Banach algebra, i.e., a Banach aly
=
Tx
=
(tk,n n )bk
phisms U : X1 X2 and W : Y1 Y2
gebra over the complex field C that is assok=1 n=1
such that T1 = W1 T2 U and T2 =
PN
ciative and contains an identity element 1
WT1 U1 .
where x = n=1 n en .
with respect to vector multiplication such
Let X1 and X2 be isomorphic linear For given bases
E and B, the matrix T is
that k1k = 1 is simply referred to as a comspaces. The linear operators T1 : X1
uniquely determined by T.
plex Banach algebra.
X1 and T2 : X2 X2 are said to be sim- Conversely, any K N matrix T defines
ilar if there exists an isomorphism U :
a linear operator with respect to given
X1 X2 such that T1 = U1 T2 U
bases forX and Y.
and T2 = UT1 U1 .
Two matrices that represent a linear operSome Important Linear Spaces
Let T : V V be a linear operator
ator on a finite-dimensional normed space
and M V a linear subspace of V such
Euclidean Space [6]. The N -dimensional Lebesgue Space [3, 7]. Let X be an arbirelative to two different bases are similar.
that T(M) M; then M is called invaricomplex Euclidean space
trary set, F the -algebra of subsets of X ,
o
n


ant under T . In this case, the restriction Linear Functionals [1, 2]. A linear funcand a nonnegative measure on F . The
CN := x : x = x0 x1 xN 1 T, xn C Lebesgue space
of T to M is a mapping of M into itself. tional is a linear operator f : V F, defined
n

Let T : H H be a linear operator on on some linear space V, whose range is in

with inner product
Lp (X , F , ) := f : X C
a Hilbert space H. If some closed lin- the scalar field F of the linear space.
N
1
X
Z
ear subspace M H and its orthogonal Hahn-Banach Theorem: Let V be a real

o
hx, yi :=
xn yn
p
or complex linear space, and let g be a
complement M are invariant under T,
measurable,
|f
|
d
<

n=0
real-valued functional on V that is subthen M is said to reduce T.
and corresponding induced norm is a finite- with norm
additive, i.e., g(u + v) g(u) + g(v) for
Any operator that maps a Banach space
Z
1/p
dimensional Hilbert space.
all u, v V, and that satisfies g(u) =
onto another Banach space is an open
p
kf
k
:=
|f
|
d
|| g(v) for every scalar . Let f be
mapping.
p
Sequence Space [3]. The sequence space
n
a linear functional, defined on a subInverse Operator. Let T : V Z be
is a Banach space for 1 p < .
space Z of V, that satisfies |f (z)| g(z)
lp := x : x = {xn }
n=0 ,
operator. Then, the inverse oper For p = , the space L (X , F , ) with a linear
for all z Z. Then, f has a linear exten1

o
X
-essential supremum norm kf k := ator T : R(T) D(T) exists iff Tv = 0

p
sion f from Z to V that satisfies |f (v)|
implies that v = 0.
xn C,
|xn | <
ess supt |f (t)| is also a Banach space.
g(v) for all v V.
1
n=0
If is finite and X = (a, b], the spaces If T exists, it is a linear operator.

The
codimension of N (f ) is 1.
1
p
p
with norm
L () := L ((a, b], F , ) are nested: If dim D(T) = N < and T exists, If A is any subspace of V with N (f ) A
!
1/p
then dim R(T) = dim D(T).

Lp () Lq () for p q.
X
and N (f ) 6= 0, then A = V.
p
2

An invertible linear operator is a homeo The spaceR L (X , F ) with inner product

kxkp :=
|xn |
For some linear functional f and some
morphism.
hf, gi := X f g d and induced norm is a
n=0
scalar , the set {v V : f (v) = } is
Let T : X Y and S : Y Z be
is a Banach space for 1 p .
Hilbert space, called the Hilbert function
called the hyperplane in V determined
bijective linear operators, where X , Y,
space. The elements of L2 (X , F , ) are
by f and .
For p = , the norm is the supremum
and Z are linear spaces. Then, the inequivalence classes of functions that differ
norm: kxk := supn |xn |.
1
verse (ST) : Z X of the composition Algebraic Dual [1]. The set V ? of all linon null sets.
An important subspace of l is the space
(also
called product) ST := S T exists ear functionals defined on a linear space V
whose elements are sequences that decay The space L1 (R) L2 (R) is a Hilbert
1
and
(ST)
= T1 S1 .
is itself a linear space, called the algebraic
to zero, i.e., xn 0 as n .
space. It is a dense subspace of L2 (R).
A bounded bijective operator T : X Y dual space of V. Its vector sum is defined
2
For p = 2, the space l with inner product For X = {0, 1, . . . , N 1} or X = Z+
between two Banach spaces X and Y has as s(v) = (f1 + f2 )(v) := f1 (v) + f2 (v) for

and the counting measure on the colX

a bounded inverse.
all v V, and the product of a scalar and

lection F of all subsets of X , the space Von Neumann Theorem: Let T : B B a vector, i.e., a functional f V ? , is defined
xn yn
hx, yi :=
2
N
2
L (X , F ) reduces to C or l , respecn=0
be a bounded operator on a Banach for all v V ? as p(v) = (f )(v) := f (v).
tively.
1/2
and norm kxk2 := hx, xi
is an infinitespace B that satisfies kI P
Tk < 1. Then, Let V be an N -dimensional linear space,

1
n
dimensional Hilbert space, called the When is a probability measure, i.e.,
T
is
invertible,
and
T
=
and let E = {e1 , . . . , eN } be a basis
n=0 (I T) .
(X
)
=
1
for
arbitrary
X
,
then
Hilbert sequence space.
1
for V. Define the set of linear functionFurthermore,
kT
k

1/(1

kI

Tk).
2
L (X , F , ) is the space of all random
als B := {f1 , . . . , fN } with fk (en ) = kn .
Space of Continuous Functions.
variables with finite second moment.
Projections [2].
A linear operator P :
Then
B is a basis for the algebraic dual
Let C[a, b] denote the space of all complexX X that satisfies P2 = P is called a
space V ? of V, and dim E = dim B; B is
valued continuous functions f : [a, b] C Schwarz Space [7]. The Schwarz space S projection.
called the dual basis of E.
with pointwise addition and scalar multipli- is the space of all infinitely differentiable,
rapidly decaying functions of a real parame- Range R(P) and null space N (P) are
cation.
C [a, b], endowed with the supremum ter t:
n
dn f (t)
norm kf k := supatb |f (t)|, is a BaLinear Functionals on Normed Spaces
S := f : R C : lim tm
=0
nach space.
t
dtn

Linear Functionals [1, 2].
Let f :
Theorem implies that every bounded lin Endowed with the inner product hf, gi :=
m, n N
Rb
U

F
be
a
linear
functional
on
a
normed
ear functional f on a subspace S U has
f (t)g (t)dt and the induced norm, this
space U.
a
a linear extension f on U that has the
Paley-Wiener
Space.
to
write
space is an inner product space but not
same norm, 
The norm kf k of a linear functional f

a Hilbert space.
Sobolev Space. to write
is the usual operator norm: kf k =
 
sup
f
(u)

 = sup |f (s)| .
supuU ,u6=0 |f (u)|.
uU ,kuk=1
sS,ksk=1
A bounded linear functional is a linear
Let U be a normed space and let u
function f that satisfies kf k a for
U. Then, there exists a bounded linsome a R.
ear functional f on U such that kf k = 1
On a normed space U, the Hahn-Banach
and f (u) = kuk.

 For every u in a normed space U,

Sesquilinear Form [1, 3.8].
Let V
and Z be liner spaces over the same scalar
|f (u)|
kuk = sup
.
field F. A sesquilinear form, or sesquilin0
kf k
f U
ear function f on V Z is a mapping f :
f 6=0
V Z F such that for all v, v1 , v2 V
Given a linearly independent set
and z, z1 , z2 Z and all scalars and
0
{f1 , . . . , fN } U , there are elements
f (v1 + v2 , z) = f (v1 , z) + f (v2 , z),
u1 , . . . , uN in U such that fj (uk ) = jk .
f (v, z1 + z2 ) = f (v, z1 ) + f (v, z2 ),
f (v, z) = f (v, z),
Convergence [1, 4.9]. For linear func f (v, z) = f (v, z).
tionals, strong and weak convergence are
Dual Space [1].
Let U be a normed equivalent, so that a sequence {fn } of
space. Then the set of all bounded lin- bounded linear functionals on a normed
ear functionals on U constitutes a normed space U is said to be
convergent if there is an f
space under the usual operator norm kf k = strongly
0
supuU ,kuk=1 |f (u)|. This space is called the
U , called the strong limit of {fn }, such
0
that kfn f k 0; this is written as fn
dual space U of U.
f;
0
0
The dual space U of a normed space U
weak convergent if there is an f U ,
is a Banach space, whether or not U is

called the weak limit of {fn }, such

complete.
that fn (u) f (u) for all u U; this
w

is written as fn f .

Linear Operators on Normed and Banach Spaces

Continuity [2, 1]. Let X and Y be normed addition (T1 +T2 )u := T1 u+T2 u, for all u
spaces, and let T : X Y be a linear oper- U, and scalar multiplication (T)u := Tu
ator.
with F.
The operator T is continuous iff
The linear space G(U, Z) is a normed

X
 X
space, whose norm is the usual operator
T
n xn =
n T(xn )
norm kTk for all T G(U, Z).
Let B ba a Banach space; then, G(U, B)
n=1
n=1
P
is a Banach space.
for every convergent series n=1 n xn
Let H is a Hilbert space, then G(H, H) is
in X .
a Banach algebra.
If T is continuous at a single point, it is
continuous.
Convergence [1, 4.9]. Let U and Z be
The linear operator T is continuous iff it normed spaces. A sequence {Tn } of operais bounded.
tors Tn G(U, Z) is said to be
If a linear operator T is continuous, it is
uniformly operator convergent if {Tn }
uniformly continuous.
converges in the operator norm
If X is finite dimensional, then T is conon G(U, Z), i.e., kTn Tk 0;
tinuous.
strongly operator convergent if {Tn u} converges strongly in Z for every u U,
Operator Norm [1, 2.7]. Let T : U Z
i.e., kTn u Tuk 0 for all u U;
be a linear operator that maps a normed
space U into a normed space Z. The opera- weakly operator convergent if {Tn u} converges weakly in Z for every u U,
tor norm is defined as
i.e., |f (Tn u) f (Tu)| 0 for all u U
kTuk
kTk := sup
and all bounded linear functionals f on U,
uU kuk
0
u6=0
that is, for all f in the dual space U of U.
where the norms on the RHS are vec- Uniform convergence implies strong convertor norms in Z and U. If D(T) = {0}, gence, which in turn implies weak converthen kTk := 0.
gence, all with the same limit.
The operator norm kTk of T is equivalent Let Tn G(B, U), where B is a Banach
to
space and U a normed space. If {Tn } is
kTk = sup kTuk.
strongly operator convergent with limit T,
uU
then T G(B, U).
kuk=1
A sequence {Tn } of operators in G(B, Z),
Bounded Linear Operators [1, 2.7].
where B and Z are Banach spaces, is
The linear operator T : U Z that maps a
strongly operator convergent iff (i) the
normed space U into a normed space Z is
sequence {kTn k} is bounded, and (ii) the
said to be bounded if there is a real number a
sequence {Tn b} is Cauchy in Z for evsuch that kTk a.
ery b in a total subset of B.
A linear operator T is bounded iff it is
Adjoint Operator [1, 4.5]. Let X and Y
continuous.
If a normed space U is finite dimen- be normed spaces and let T : X Y be
for any
sional, then every linear operator on U is a bounded linear operator. Then,
0
bounded
linear
functionals
f

X
and
g
bounded.
0
0
0
T = 0 iff hTu, zi = 0 for all u U Y , the adjoint operator T : Y X of T
is defined by f (x) = (T g)(x) = g(Tx) for
and z Z.
all x X .
The null space N (T) of T is closed.
If {un } a sequence in D(T), then un u The adjoint operator T is linear and
implies Tun Tu.
bounded, and kT k = kTk.
For bounded linear operators T1 : X If T is represented by a matrix T, then
Y and T2 : Y Z on normed
the adjoint operator T is represented
spaces X , Y, and Z, it follows that
by TT .
kT1 T2 k kT1 kkT2 k, and for T : X X
Let S : X Y be another bounded linear
that kTn k kTkn .
operator. Then
Uniform
Boundedness
Theorem:
(S + T) = S + T .
Let {Tn } be a sequence of linear op (T) = T , F.
erators Tn : B U from a Banach
(ST) = T S .
space B into a normed space U such
1
exists and T1 B(X , Y),
that kTn bk cb < for every b B If T
1
then (T )
also exists, (T )1
and every n = 1, 2, . . . . Then, the se0
0
quence of norms {kTn k} is bounded, i.e.,
B(X , Y ), and (T )1 = (T1 ) .
there exists a c such that kTn k c for
Closed Linear Operators [1, 4.13].
all n = 1, 2, . . . .
A bounded linear operator T from a Ba- Let U and Z be normed spaces and let T :
nach space B onto a Banach space Z has D(T) Z be a linear operator with dothe property that the image T(B1 (0)) of main D(T) U. Then, T is called a
the open unit ball around the origin con- closed linear operator if its graph G(T) :=
{(u, z) : u D(T), z = Tu} is closed in the
tains an open ball around 0 Z.
Open mapping theorem: A bounded linear normed space U Z.
operator T from a Banach space onto an- Closed graph theorem: Let T be a closed
operator. If D(T) is closed in V, the opother Banach space is an open mapping.
erator T is bounded.
Hence, if T is bijective, T1 is continuous
Let T : D(T) Z be a linear operator,
and thus bounded.
where D(T) U and U, Z are normed
Operator Topologies [1, 2]. Let G(U, Z)
spaces. Then, T is closed iff it has the foldenote the set of all bounded linear operalowing property: If un u for un
tors from a normed space U into a normed
D(T), and Tun z, then u D(T)
space Z over the same scalar field. The
and Tu = z.
set G(U, Z) is a linear space under operator

Compact Linear Operators [1, 2].

Let U and Z be normed spaces. A linear
operator T : U Z is called compact or
completely continuous if for every bounded
subset S U, the image T(S) is relatively
compact, i.e., the closure T(S) is (sequentially) compact.
Every compact linear operator T is
bounded and, therefore, continuous.
If dim U = , the identity operator I,
which is continuous, is not compact.
A linear operator T : U Z is compact
iff it maps every bounded sequence {un }
in U onto a sequence {Tun } in Z that
has a convergent subsequence.
If T is bounded and dim R(T) < ,
then T is compact.
If U is a finite-dimensional normed linear
space, every linear operator defined on U
is compact.
Given  > 0, there exists a finitedimensional subspace M R(T) such
that
inf kTu mk < kuk

for any u U.
Let {un } be a weakly convergent sew
u. Then {Tun }
quence in U with un
is strongly convergent in Z and has the
strong limit z = Tu.
If T is compact, so is its adjoint opera0
0
tor T : Z U .
Let {Tn } be a sequence of compact linear
operators from a normed space U into
a Banach space B. If {Tn } is uniformly
operator convergent, i.e., kTn Tk 0,
then the limit operator T is compact.
A compact linear operator T : U B
from a normed space U into a Banach
:
space B has a compact linear extension T

U B, where U is the completion of U.

Let T : B A and S : B A be compact linear operators, where B and A are
Banach spaces. Then, T + S is compact.
Let T : U U be a compact linear operator and S : U U a bounded linear
operator on a normed space U. Then TS
and ST are compact.

mM

Spectral Theory of Linear Operators

Resolvent, Spectrum [2, 1].
Let B
be a complex Banach space, and let T :
T one-to-one?
RRR
D(T) R(T) be a linear operator
RRR no
RRR
yes
with D(T), R(T) B.
RRR
R(

Associated with T is the the opera Sp (T)
R(T
)
dense
in
X
?

tor T := T I, where C and I

QQQ
QQQ no
denotes the identity operator.
QQQ
yes
QQQ
If T has an inverse defined on its range,
Q(

it is called the resolvent of T and denoted
1
Sr (T)
R
(T)
continuous?
1

as R (T) := T = (T I) on R(T ).
QQQ
QQQ no
QQQ
The resolvent set Q(T) of T is defined as the
yes
QQQ
Q(
set of all complex numbers such that the

range of T is dense in B and that T has
Q(T)
Sc (T)
a continuous inverse defined on its range.
The numbers Q(T) are called regular The four sets are pairwise disjoint
values. The set S(T) := Q(T)c is called the
and C = Q(T) Sp (T) Sc (T) Sr (T);
spectrum of T; a S(T) is called a specsome of the sets may be empty.
tral value of T. The spectrum S(T) can be If R (T) exists, it is a linear operator.
partitioned into three disjoint sets:
Let B be a complex Banach space, T :
B B a linear operator, and Q(T).
The point spectrum Sp (T) is the set such
If T is closed or bounded, then, R (T)
that T is not one-to-one. A Sp (T)
is defined on the whole space B and is
is called an eigenvalue of T.
bounded.
The continuous spectrum Sc (T) is the
set such that T is one-to-one, has its
Eigenvalues [1, 7]. Let U be a normed
range dense set in B, but R (T), defined
space over the complex field and T : D(T)
on R(T ), is not continuous and, thereU a linear operator with domain D(T) U.
fore, unbounded.
The residual spectrum Sr (T) is the set The resolvent R (T) exists iff Tu = 0
implies u = 0, i.e., the null space N (T)
such that T is one-to-one, but R(T ) is
is {0}.
not dense in B.
If T u = 0 for some u 6= 0, then
In summary:
Sp (T). The vector u is then called an
eigenvector of T with eigenvalue .
The subspace of D(T) that consists of 0
and all eigenvectors of T with eigenvalue is called the eigenspace of T corresponding to that eigenvalue.
Eigenvectors with different eigenvalues
constitute a linearly independent set.

Spectral Properties of Operators on Normed Spaces

Bounded Linear Operators on a Complex Banach Space [1, 7.3]. Let B be a
complex Banach space, and let T G(B, B)
be a bounded linear operator.
The resolvent set Q(T) is not empty.
The spectrum S(T) is not empty.
The resolvent set Q(T) is open; hence,
the spectrum S(T) is closed.
If kTk < 1, then (I T)1 exists,
is a bounded linear operator on the
whole space B, and has the following series expansion, convergent in the norm
on G(B, B):

X
(I T)1 =
Tn = I + T + T2 + . . .
n=0

and k(I T)1 k (1 kTk)1 .

For every 0 Q(T), the resolvent R (T)
has the representation

X
R (T) =
( 0 )n Rn+1
0 .
n=0

The resolvent R (T) is holomorphic at

every point 0 of the resolvent set Q(T).

Hence, it is locally holomorphic on Q(T).
The spectral radius of T is defined as rT :=
supS(T) ||.
The spectral radius is given as rT =
limn kTn k1/n .
The spectrum S(T) is compact and lies
in a disk with spectral radius rT kTk.
Let , R (T). Then,
The resolvent R (T) satisfies the
Hilbert relation, also called resolvent
identity:
R R = ( )R R ;
R (T) commutes with any S G(B, B)
that commutes with T;
R R = R R .
Spectral mapping: Let p() := n n +
n1 n1 + + 0 0 with n 6= 0.
Then, S(p(T)) = p(S(T)). That is, the
spectrum of the operator p(T) = n Tn +
n1 Tn1 + + 0 I consists of all those
values that the polynomial p assumes on
the spectrum S(T) of T.

 There exists a smallest integer n = r, Normal Operators [2]. Let T : H H

depending on , such that
be a normal operator of a Hilbert space H
r+1
r+2
r
into itself.
N (T ) = N (T ) = N (T ) . . .
Let en be an eigenvector associated with
and
the eigenvalue n . Then, the vector en is
r+1
r+2
r
T (U) = T (U) = T (U) . . .
also an eigenvector of the Hilbert adjoint
If r > 0, the inclusions
operator T? of T and associated with the
N (T0 ) N (T1 ) N (Tr )
eigenvalue n .
The null space satisfies N (T I) =
and
N (T? I).
T0 (U) T1 (U) Tr (U)
For any 6= , the null spaces N (T I)
are proper. Furthermore, the space U can
and N (T I) are orthogonal to one anr
r
be represented as U = N (T ) T (U).
other.
For each complex number , the closed
linear subspace N (T I) reduces T.
and Functionals on Hilbert Space
kT2 k = kTk2 .
A bounded linear operator T on a Hilbert
space H is normal iff kT? hk = kThk for
every h H.
The residual spectrum Sr (T) of a normal
normal operators
operator is empty.
!
!

A
complex number is in S(T) iff
TT = T T
there exists a sequence {hn } with hn
H, khn k = 1 for all n, such that k(T
I)hn k 0 as n ; in other words,
self-adjoint operators
the operator T I is not bounded below.
!

Let
a bounded linear operator H on
T=T
I
a Hilbert space H have the Cartesian
unitary
decomposition H = T + iS, where T
operators
and S are self-adjoint. Then, H is nororthogonal
mal iff T and S commute.
In that
!
1

projections U : U = U
case, max kTk2 , kSk2 kHk2 kTk2 +
P
kSk2 .
O
Let T and S be normal operators on
trace-class operators
a Hilbert space H such that one commutes with the adjoint of the other, i.e.,
TS? = S? T and T? S = ST? , or such
Hilbert-Schmidt operators
that the two operators commute, i.e.,
TS = ST; then, T + S, TS, and ST are
compact operators
normal.

Compact Linear Operators [1, 8].

Let T : U U be a compact operator on a
normed space U, and let T := T I.
Every spectral value S(T), 6= 0, if
it exists, is an eigenvalue of T.
The set of eigenvalues Sp (T) is at most
countable, and its only possible limit
point is = 0.
If = 0 Q(T), then T is finite dimensional.
For every 6= 0 and every n = 1, 2, . . . ,
the null space N (Tn ) is finite dimensional
and the range R(Tn ) is closed.

Linear Operators

continuous
linear
operators

T:HH

Representation of Functionals [1,

3.8].
Riesz Theorem: Every bounded linear
functional f on a Hilbert space H can be
represented by an inner product f (h) =
hh, zi, where h H, and where z H
is uniquely determined by f and has
norm kzk = kf k.
Riesz representation: Let H1 and H2 be
Hilbert spaces, let h1 H1 , h2 H2 ,
and g : H1 H2 F a bounded
sesquilinear form. Then g has a representation g(h1 , h2 ) = hSh1 , h2 i, where S :
H1 H2 is a bounded linear operator
that is uniquely determined by g and has
norm kSk = kgk.

A bounded linear operator T : H H on a

Hilbert space H is said to be
normal, if TT? = T? T,
unitary, if T is bijective and if T? = T1 ,
self adjoint or Hermitian, if T? = T.
If T is self adjoint or unitary, it is normal.

Unitary Operators.
Let the operators U, V : H H be unitary, H a Hilbert
space. Then,
U is isometric, i.e., kUhk = khk for
all h H,
kUk = 1,
U1 is unitary,
UV is unitary.
A bounded linear operator on a Hilbert
space over the complex field is unitary iff
Hilbert Adjoint Operator [1, 2]. Let T :
it is isometric and onto.
H Z be a bounded linear operator
that maps the Hilbert space H into the Polar Decomposition [8, 30]. Let T :
Hilbert space Z. The Hilbert adjoint op- H H be a compact linear operator on a
erator T? of T is the operator T? : Z H separable complex Hilbert space H; let T?
such that hTh, zi = hh, T? zi for all h denote the Hilbert adjoint of T.
H and z Z. This operator exists, is The operator T can be factored as T =
unique, and is a bounded linear operator
UA, where A is a positive Hermitian opwith norm kT? k = kTk.
erator and U? U = I on the range of A.
Let S : H Z be another bounded linThe above factorization is called the poear operator, and let be any scalar. The
lar decomposition of T; the operator A is
Hilbert adjoint operator has the following
called the absolute value of T. The polar
properties:
decomposition exists even if T is bounded
I? = I.
instead of compact.
hT? h, zi = hz, Thi.

The absolute value A can be taken

(S + T)? = S? + T? .
as A := (T? T)1/2 , the unique positive
(T)? = T? .
square root of T? T; the operator U satis (T? )? = T.
fies U : Ah Th for all h H.
kT? Tk = kTT? k = kTk2 .
If T is compact, then its absolute value A
T? T = 0 T = 0.
is compact.
(ST)? = T? S? .
If T can be represented by a matrix T, Singular Values [8, 30]. Let T : H H
be a compact linear operator on a separable
then T? can be represented by TH .
?
Let T : H H a bounded linear operator complex Hilbert space H, and let T be its
Hilbert
adjoint.
Furthermore,
let
T
=
UA be
that maps a Hilbert space H into itself.
the polar decomposition of T, and let {n }
?
The ranges and null spaces of T and T denote the set of nonzero eigenvalues of A;
are related as follows:
they are all positive, as A is Hermitian. Let
R(T) = N (T? ).
the n be indexed in decreasing order. The
N (T) = R(T? ).
numbers n are called the singular values
Let T be continuous, and let M be a of T, denoted also as n (T).
closed linear subspace of H. Then, M The singular values of T form an at most
is invariant under T iff M is invariant
countable sequence whose only possible
under T? .
limit point is 0.
A closed linear subspace M H re- Let the nonzero eigenvalues of T
duces T iff M is invariant under both T
be 1 (T), 2 (T), . . . , arranged in decreasand T? .
ing order of their absolute value, includ The Hilbert adjoint operator T? : Z H
ing multiplicities. Then, for any N Z+
0
0
and the adjoint operator T : Z H
N
N
Y
Y
are related as T? = A1 T A1
,
where
A
:
1
|
(T)|

n (T).
2
n
0
0
H H and A2 : Z Z are bijective,
n=1
n=1
isometric, conjugate linear operators that
are uniquely defined by Rieszs theorem.

Bounded Self-Adjoint Linear Operators [1, 2]. Let T : H H be a bounded

self-adjoint linear operator on a complex
Hilbert space H, let T := T I, and
let h H.
The set of all self-adjoint linear operators
on H is a closed set in G(H, H).
The set of all self-adjoint linear operators
on H forms a real normed linear space
under the operator norm.
A bounded linear operator T on a complex Hilbert space H is self adjoint
iff hTh, hi = hh, Thi is real for all h H.
If H is a real Hilbert space, the direct
part holds but the converse is no longer
true.
The spectrum S(T) of T lies in the closed
intervall [mT , MT ] R, where
mT = inf hTh, hi,
MT = suphTh, hi.
khk

khk

Both mT and MT are spectral values of T.

The operator norm of T is given by
kTk = max(|mT | , |MT |) = sup |hTh, hi| .
khk

Eigenvectors that correspond to numerically different eigenvalues of T are orthogonal

A number belongs to the resolvent
set R (T) iff there exists a c > 0 such
that kT hk ckhk for every h H.
The product of two self adjoint linear operators on a Hilbert space is self adjoint
only if the operators commute.
Every bounded linear operator T : H
H has a so-called Cartesian decomposition: T = A + iB, where A and B are
self-adjoint.
The Cartesian decomposition is
unique.
A = 1/2(T + T? ).
B = 1/(2i)(T T? ).
If is an eigenvalue of T, then =
+ i, where is an eigenvalue of A
and is an eigenvalue of B.
Nonnegative Self-Adjoint Linear OpConsider the set of
erators [1, 9].
all bounded, self-adjoint, linear operators
on a complex Hilbert space H. A reflexive partial ordering  on this set is defined by T1  T2 iff hT1 h, hi hT2 h, hi
for h H. A bounded, self-adjoint, linear operator T is said to be nonnegative
(although not strictly correct, sometimes
also called positive) and denoted T  O,
if hTh, hi 0 for all h H.
T1  T2 O  T2 T1 .
If two bounded, self-adjoint, linear operators T and S are nonnegative and commute, i.e., TS = ST, then their prod-

uct ST is nonnegative.
If T is bounded and self adjoint, then T2
is nonnegative.
A monotone sequence {Tn } of bounded,
self-adjoint, linear operators is a sequence
that is either monotonically increasing, i.e.,
T1  T2  T3  . . . , or monotonically decreasing, T1  T2  T3  . . . .
Let {Tn } be a monotonically increasing
sequence of bounded, self-adjoint, linear
operators such that T1  T2  . . . 
Tn  . . .  S, where S is also bounded
and self adjoint. Suppose that all elements of the sequence commute pairwise
and also commute with S. Then, {Tn } is
strongly operator convergent, Tn h Th
for all h H, and the limit operator T
is linear, bounded, self adjoint, and satisfies T  S.
A bounded, self-adjoint linear operator S
is called a square root of another bounded,
self-adjoint, linear operator T if S2 = T.
If, in addition, S  O, then S is called a
nonnegative square root of T and is denoted
by S = T1/2 .
Every nonnegative, bounded, self-adjoint,
linear operator T : H H on a complex
Hilbert space H has a nonnegative square
root S that is unique.
The square-root operator S of T commutes with every bounded linear operator
on H that commutes with T.
Compact Normal Operators [2].
Let T : H H be a normal operator on a
nontrivial Hilbert space H, and let T have
the Cartesian decomposition T = A + iB.
The operator T is compact iff both A
and B are compact.
The operator T is compact iff T? is compact.
If T is compact, it has an eigenvalue
with max{kAk, kBk} ||. If T is
self-adjoint, then it has an eigenvalue
with = kTk.
If T is compact and has no eigenvalues,
then H = {0}.
If H is not separable, then = 0 is necessarily an eigenvalue of any compact normal operator on H.
Hilbert-Schmidt Operators [9, 2, 8].
Let {xn } be an orthonormal basis for a
Hilbert space H. A bounded linear operator T : H H isPcalled a Hilbert
Schmidt(HS) operator if n=1 kTxn k2 < .
The number

X
1/2
kTkHS :=
kTxn k2
n=1

is called the Hilbert-Schmidt norm of T.

The HS norm does not depend on the
choice of orthonormal basis for H.
The HS norm of a matrix is also called
the Frobenius norm.
If T is HS, then T? is HS, and kTk
kTkHS , as well as kTkHS = kT? kHS .
Every HS operator is compact; hence it
is bounded and continuous.
Every HS operator is the limit in HSnorm of a sequence of operators with
finite-dimensional range.
A P
compact linear operator is HS
iff n n2 (T) < .
For a representation of a given Hilbert
space as L2 (H, M , ) with positive measure and the corresponding collection M of measurable subsets, HS operators are those operators T that have a
representation in
Z the form
(Tf )(t) =

k(t, s)f (s)d(s),

H

where f = f (t) L2 (H, M , ), and the

integral kernel k(t, s) satisfies
ZZ
2
|k(t, s)| d(s)d(t) < .
H

If T S(H), and f is a single-valued analytic function on S(T) that vanishes at

zero, then f (T) is a HS operator, and the
mapping T f (T) of S(H) into itsef is
continuous.
The set S(H) of all HS operators on a
Hilbert space H, together with the HS norm,
is a Banach algebra with kTSkHS kTkHS
kSkHS for every T, S S(H). It contains
operators of finite range as a dense subset.
The set of HS operators is a self-adjoint
ideal in G(H, H), the Banach algebra of all
bounded linear operators in Hilbert space.

Trace Class Operators [8, 30]. A comas the product of two HS operators.
pact linear operator T : H H on a Hilbert Let {xn } be any orthonormal basis of H.
space H is said to be in trace class if
For a trace class operator T, the trace is

X
defined as the limitX
of the series
n (T) < .
tr T :=
hTxn , xn i.
n=1

The above sum defines the trace norm kTktr : This series converges absolutely. For trace

X
class operators
P T
kTktr :=
n (T).
tr T = n n (T), where n (T) are the
n=1
eigenvalues of T.
For T in trace class and any bounded oper- If T is a trace class operator that has no
ator B : H H
eigenvalues except = 0. Then, tr T = 0.
kTk kTktr ,

|tr T| kTktr .
kTktr = kT? ktr ,
tr T is a linear mapping of T.
kBTktr kBktr kTktr ,
tr T? = (tr T) .
kTBktr kBktr kTktr .
For any bounded operator B, tr(TB) =
For any pair of trace class operators T
tr(BT).
and S, T+S is trace class, and kT+Sktr Let T be a trace class operators, and
kTktr + kSktr .
let {Tn } be a sequence of degenerate op The trace class is a two-sided ideal in the erators, i.e., operators with finite range that
algebra of all bounded linear operators converge to T in trace norm. Then, the
on a complex Hilbert space.
determinant det(I + Tn ) of the matrix rep Trace class operators form a Banach resentation of I + T , I + T , tends to a
n
n
space with respect to the trace norm.
limit that is independent of the choice of
Every trace class operator is HS.
the sequence {Tn }. This limit is called the
The product of two HS operators T and S determinant of I + T:
is in trace class, and kSTktr kSkHS
det(I + T) := lim det(I + Tn ).
kTk .
HS

Every trace class operator can be written

The Spectral Theorem

Orthogonal Projection [2, 1]. A projection P : H H on a Hilbert space H is
called an orthogonal projection if its range
and null space are orthogonal: R(P)
N (P).
A bounded linear operator P : H H on
a Hilbert space H is an orthogonal projection if P is self adjoint and idempotent,
i.e., P2 = P.
An orthogonal projection is continuous
(even if H is not complete).
A continuous projection on a Hilbert
space is orthogonal iff it is self-adjoint.
P  O;
kPk 1 with equality if P(H) 6= {0}.
N (P) = R(P) and R(P) = N (P) .
For any orthogonal projection P on a
Hilbert space H and for any h H,
hPh, hi = kPhk2 .
Each h H can be written uniquely
as r + n, where r R(P) and n N (P);
furthermore, kxk2 = krk2 + knk2 .
Let M be any closed subspace of a Hilbert
space H. Then there is exactly one orthogonal projection P with R(P) = M.
Let {en } be a countable orthonormal set
in H such that M = span{en }; then, the
mapping P : H H defined by
X
Ph :=
hh, en ien

The sequence {Pn } is strongly operator

convergent, say Pn h Ph for all h H,
and the limit operator P is a projection
on H.
The limit operator P projects H onto

[
P(H) =
Pn (H).
n=1

The limit operator P has the null space

\
N (P) =
N (Pn ).
n=1

Spectral Family [1, 2]. A real spectral

family, is a collection E := {E : R}
of projection operators E on a Hilbert
space H of any dimension that satisfy the
following properties for any h H:
E E for ( ); hence, E E =
E E = E .
lim E h = 0.
lim E h = h.
E+ h := lim E h = E h.
Special cases:
A countable resolution of the identity is
a sequence {Pn } of orthogonal projection
operators with
P Pn Pm = O for n 6= m
so that I =
n Pn , where the sum is
strongly operator convergent. The sequence {Pn } defines a spectral family E
n
with
X
for any h H is the orthogonal projecE :=
Pn .
tion of H onto M.
n
Let M be a closed subspace of H, let h A spectral family on an interval [a, b] R
H, and let P be the orthogonal projection
is a spectral family E that satisfies E =
on H with R(P) = M. Then, khPhk =
O for < a and E = I for b.
inf mM kh mk.
Let P1 and P2 be orthogonal projections Compact Normal Operators [2].
on a Hilbert space H, and let Y1 := P1 (H) Let T : H H be a compact normal
operator on a Hilbert space H.
and Y2 := P2 (H).
The composite operator P := P1 P2 is a There is a countable resolution of the
identity {Pn } and a sequence of complex
projection on H iff P1 and P2 commute.
numbers {n } such that
In this case, P projects H onto Y =
X
Y1 Y2 . Conversely, the projection
T=
n Pn ,
onto span{Y1 , Y2 } is P1 + P2 P1 P2 .
n
The sum P := P1 + P2 is a projection
where convergence is uniform in the opoperator iff Y1 Y2 . In this case, P
erator norm.
projects H onto Y1 Y2 .
There exists an orthonormal basis {en }
The difference P := P2 P1 is a projecof eigenvectors and a corresponding setion on H iff Y1 Y2 . In this case, P
quence of
P eigenvalues {n } such that,
projects H onto the orthogonal compleif h =
n hh, en ien is the Fourier exment of Y1 in Y2 .
pansion for h H, then
X
The following conditions are equivalent
Th =
n hh, en ien .
P 2 P1 = P1 P2 = P1 ,
n
P
Y1 Y 2 ,
A weighted sum of projections n n Pn ,
N (P1 ) N (Y2 ),
where {Pn } is a resolution of the iden kP1 hk kP2 hk for all h H,
tity, and {n } is a sequence of com P 1  P2 .
plex numbers, is compact if (i) for every
Let {Pn } be a monotonically increasing
nonzero n , the range of Pn is finite disequence of projection operators Pn on a
mensional, and (ii) for every real > 0,
Hilbert space H. Then,
the number of n with |n | is finite.

Bounded Self-Adjoint Operators [1,

fine f (T) as the limit p(T) of the polyno9]. Let T : H H be a bounded, selfmial Tn := pn (T) := n Tn + n1 Tn1 +
adjoint, linear operator on a Hilbert space H,
+ 0 I for n , where pn () is
let T := T I, and define the positive part
such that it converges uniformly to f ()
2 1/2
on [mT , MT ]. Then, the operator f (T)
of T as T+
+ T )/2. Further := ((T )
has the spectral representation
more, let Y := N (T+
)
denote
the
null
space

Z MT
+
of T .
f (T) =
f ()dE ,
Let E with R be the projection of H
mT 0
+
onto the null space Y of T . Then, the
and for all x, y H,
Z MT
collection E (T) := {E : R} is the
unique spectral family associated with T
hf (T)x, yi =
f ()dw().
mT 0
on the interval [mT , MT ] R.
For < , the projection operator E
The operator f (T) is self adjoint.
E satisfies (E E )  T(E E ) 
If f () = f1 ()f2 (), then f (T) =
(E E ).
f1 (T)f2 (T).
The mapping E has a discontinuity
If f () 0 for all [mT , MT ], then
at 0 , i.e., E0 6= E+ , iff 0 is an eigenf (T)  O.
0
value of T. In this case, the eigenspace
If f1 () f2 () for all [mT , MT ],
that corresponds to the eigenvalue 0
then f1 (T)  f2 (T).
is N (T 0 I) = (E0 E+ )(H).
kf (T)k max[mT ,MT ] |f ()|.
0
If a bounded linear operator commutes
A real 0 belongs to the resolvent
with T, it also commutes with f (T).
set R (T) iff there is an  > 0 such
that E (T) is constant on the interval [0
Unitary Operators [1, 10.5]. Let U :
, 0 + ].
H H be a unitary operator on a complex
A real 0 belongs to the continuous specHilbert space H.
trum Sc (T) iff the mapping E is
continuous at 0 (thus E0 = E+ ) and is The spectrum S(U) is a closed subset of
0
the unit circle. Consequently, || = 1 for
not constant in any neighborhood of 0 .
every S(U).
A bounded, self-adjoint, linear operator T on There exists a spectral family E = {E }
a complex Hilbert space H has the spectral
on [, ] such that
Z
representation
Z MT
Z MT
ei dE .
U=

T=
dE = mT EmT +
dE ,
mT 0
mT
for every continuous function f on the
unit circle,
where E = {E } is the spectral family asZ


sociated with T, and the integral is to be
f (U) =
f ei dE ,
understood in the sense of uniform operator

convergence in the norm on B(H, H).

where the integral is to be understood
For x, y H,
in the sense of uniform operator converZ MT
gence.
dw(),
hTx, yi =
For all x, y H,
mT 0
Z


where w() := hE x, yi, and the integral
hf (U)x, yi =
f ei dw(),
is of Riemann-Stieltjes type.

Let f () : [mT , MT ] R be a continuous,

where w() := hE x, yi, and the integral
real-valued function on [mT , MT ]. Deis an ordinary Riemann-Stieltjes integral.

Linear Operator Equations

Fredholm Alternative [1, 8.7].
A
bounded linear operator S : U U on a
normed space is said to satisfy the Fredholm
alternative if either one of the following conditions holds
The nonhomogeneous equations
Sx = y,
S f = g
have unique solutions x and f , respectively, for every given y U and g
0
U , and the corresponding homogeneous
equations
Sx = 0,
S f = o
have only the trivial solutions x = 0
and f = 0, respectively.
The homogeneous equations
Sx = 0,
S T = 0
have the same number of linearly
independent solutions x1 , x2 , . . . , xN
and f1 , f2 , . . . , fN , respectively, and the
corresponding nonhomogeneous equations
Sx = y,
S f = g
are not solvable for all y and f , respectively. They have a solution iff y and g
are such that fn (y) = 0 and g(xn ) = 0
for all n = 1, 2, . . . , N .
For a compact linear operator T on a normed
space U, the operator T := T I, for 6=
0, satisfies the Fredholm alternative.

Let T : U U be a compact linear operator

0
0
on a normed space U and T : U U its
0
adjoint operator. For x, y U, f, g U ,
and 6= 0, consider the set of linear operator
equations
Tx x = y
(y X given) (OE1)
Tx x = 0
(OE2)
T f f = g

(g X given) (OE3)

T f f = 0.
(OE4)
Then,
Equation (OE1) has a solution x
iff f (y) = 0 for all solutions f of (OE4).
Hence, if f = 0 is the only solution
of (OE4), then (OE1) is solvable for every y.
Equation (OE3) has a solution f
iff g(x) = 0 for all solutions x of (OE2).
Hence, if x = 0 is the only solution
of (OE2), then (OE3) is solvable for every g.
Equation (OE1) has a solution x for every y U iff x = 0 is the only solution
of (OE2).
Equation (OE3) has a solution f for ev0
ery g U iff f = 0 is the only solution
of (OE4).
Equations (OE2) and (OE4) have the
same number of linearly independent solutions.
Linear Operator Equations [1, 8.5].

References

[5] , Functional Analysis, 2nd ed.

[1] E. Kreyszig, Introduction to Functional Analysis with Applications. New York, NY, [6]
U.S.A.: Wiley, 1989.
[7]
[2] A. W. Naylor and G. R. Sell, Linear Operator Theory in Engineering and Science.
New York, NY, U.S.A.: Springer, 1982.
[8]
[3] M. Pourahmadi, Foundations of Time Series Analysis and Prediction Theory. New
[9]
York, NY, U.S.A.: Wiley, 2001.
[4] W. Rudin, Real and Complex Analysis, 3rd ed.
1987.

New York, NY, U.S.A: McGraw-Hill,

New York, NY, U.S.A.: McGraw-Hill, 1991.

N. Dunford and J. T. Schwarz, Linear Operators, New York, NY, U.S.A., 1958, vol. 1.
R. L. Allen and D. Mills, Signal Analysis: Time, Frequency, Scale, and Structure.
Piscataway, NJ, U.S.A.: IEEE Press, 2004.
P. D. Lax, Functional Analysis.

New York, NY, U.S.A.: Wiley, 2002.

N. Dunford and J. T. Schwarz, Linear Operators.

1963, vol. 2.

New York, NY, U.S.A.: Wiley,