0% found this document useful (0 votes)
28 views8 pages

Chapt6 Upd 1

This chapter discusses the Riesz theorem and properties of adjoint operators between inner product spaces. The Riesz theorem states that any linear functional on a finite-dimensional vector space can be written as an inner product with a unique vector. The adjoint operator T* of a linear operator T between inner product spaces is defined such that the inner product of T*w with v equals the inner product of w with Tv. Important properties are proven, including that the kernel and range of T and T* are orthogonal complements of each other. The ranks of T and T* are also shown to be equal.

Uploaded by

kegici1148
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
28 views8 pages

Chapt6 Upd 1

This chapter discusses the Riesz theorem and properties of adjoint operators between inner product spaces. The Riesz theorem states that any linear functional on a finite-dimensional vector space can be written as an inner product with a unique vector. The adjoint operator T* of a linear operator T between inner product spaces is defined such that the inner product of T*w with v equals the inner product of w with Tv. Important properties are proven, including that the kernel and range of T and T* are orthogonal complements of each other. The ranks of T and T* are also shown to be equal.

Uploaded by

kegici1148
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 8

Chapter 6

Riesz Theorem. Adjoint operators. Ker T


and Ran T ∗. Rank Theorem for T and T ∗

In this short chapter we prove two theorems in the title, and this Chapter is mainly theory. (This
corresponds to [HC] Kap. 7.)

Notation and Terminology


• Linear functions V → K are called linear functionals; usually written as f, g etc.

• The orthogonal complement X ⊥ ⊂ V of a subspace X ⊆ V :

X ⊥ := {y ∈ V; y ⊥ X} = {y ∈ V; y ⊥ x ∀x ∈ X}

i.e., X ⊥ is the subset of all orthogonal vectors of X.

• Adjoint T ∗ of linear operator T : V → W on complex vector spaces. If V and W are real


vector space T ∗ is written also at T t (read T transpose).

Goal
• Riesz representation for linear functionals using inner product.

• Adjoint T ∗ of a linear transformation T .

• Relations between the four subspaces Ker T, Ran T ∗ , Ker T ∗ , Ran T

1
2

6.1 Riesz Theorem for finite dimensional inner product spaces


Let K = R, C be the underlying field. All vector spaces in Chapters 7-8 are inner product vector
spaces.
We are interested in finding all linear functions, called also linear functional, on vector spaces. Linear
functions on R are of the form f (x) = ax as we all know; on R2 there are of the form f (x) =
ax1 +bx2 as we can check it along one dimensional subspaces Re1 and Re2 . The linear form ax1 +bx2
on R2 can be written as hx, ui, u = [a, b]. The Riesz theorem for finite-dimensional real or complex
vector spaces is relatively easy to understand and prove.

Theorem 1. (Riesz) Let f : V → K be a linear functional on a finite dimension vector space V .


Then there exists a unique u ∈ V such that

f (v) = hv, ui, v ∈ V.

Proof. We choose {e1 , . . . , en } an ONB for V . (Such basis always exists; see Chapter 1 and Gram-
Schimdt process in Chapter 5). Let u = u1 e1 + · · · + un en where

uj := f (ej ), 1 ≤ j ≤ n.

Then for any vector v, we have (see Proposition 2 in Chapter 5),

v = v1 e1 + · · · + vn en , vj = hv, ej i, 1 ≤ j ≤ n.

Thus
f (v) = v1 if (e1 ) + · · · + vn f (en ) = v1 u¯1 + · · · + vn u¯n .
On the other hand,

hv, ui = hv1 e1 + · · · + vn en , u1 e1 + · · · + un en i = v1 u¯1 + · · · + vn u¯n

since {ej } is ONB and all products hei , ej i vanish except for i = j in which case they are 1. This
proves f (v) = hv, ui for all v ∈ V .

Exercise 6.1. Let W ⊂ R4 be the subspace

W = {x ∈ R4 ; x1 + x2 + x3 + x4 = 0}

Let f be the linear functional f : W → R, x → x1 . Find the vector u for T in Riesz Theorem. (Hint:
Follow the proof of Riesz theorem or use orthogonal projection.)
3

6.2 Adjoint operator T ∗ : W → V of T : V → W


6.2.1 Defintion and property of T → T ∗
The transpose At of a real 2 × 2-matrix A = [aij ]1≤i,j≤2 can be described as

a11 = hAe1 , e1 i = he1 , At e1 i, a12 = hAe2 , e1 i = he2 , At e1 i,

a21 = hAe1 , e2 i = he1 , At e2 i, a22 = hAe2 , e2 i = he2 , At e2 i


By the bi-linear property of the inner product this is equivalent to

hAx, yi = hx, At yi, x, y ∈ R2 .

We can now define of the adjoint operator T ∗ for an operator T : V → W on complex vector spaces.
The real vector spaces and their operators can be treated similarly.

Definition 6.1. Let V, W be a finite dimensional complex vector space with inner products. The
Hermitian adjoint T ∗ of a linear operator T : W → V is defined by

hT ∗ w, vi = hw, T vi, v ∈ V, w ∈ W.

(Note that the inner products are computed in different spaces.) For a real linear transformation T
the Hermitian adjoint T ∗ are sometimes just called adjoint or transpose and denoted by T t . We shall
use T ∗ to avoid misunderstanding.

6.2.2 The matrices of the adjoint [T ∗ ]E and [T ]E


Let E = {ej , j = 1, . . . , n} be an orthonormal basis of V and let [T ]E be the matrix of T . Then if
follows from the defintion of T ∗ that the matrix [T ∗ ]E is

[T ∗ ]E = [T ]tE ,

the conjugated transpose of [T ]E . For a complex matrix 2 × 2-matrix


 
a b
A= ,
c d
 
∗ ā c̄
A = .
b̄ d¯
4

6.3 Relations between Ker T, Ker T ∗, Ran(T ) = T V, Ran(T ∗) =


T ∗W for T : V → W between inner product spaces V → W
Recall that the conjugate At of a real m × n-matrix A is an n × m-matrix. We know from Linear
Algebra I that the ranks of At and A are the same. This result seems a very natural result, it is proved
in Linear Algebra I by using row reduction which seems less natural. We shall see now that, when
a m × n-matrix A is viewed as a linear transformation A : Rn → Rm of Euclidean spaces then its
transpose At : Rm → Rn can be expressed in terms of the inner product and not only the ranks
rank A = dim Ran A and rank At = dim Ran At are the same, but also there is more refined relation
between the spaces Ran A, Ker A, Ran At , Ker At .
Recall the rank theorem (see Lecture Note - Week 2) that if T : V → W is a linear transformation,
then
dim V = dim Ker T + rank T.
We have just defined T ∗ : W → V . In Linear Algebra I we know that for matrices rank A = rank At .
Thus we have also rank T = rank T ∗ for linear transformations. We get thus a refined version of the
rank theorem.

Theorem 2. Let T : V → W be a linear transformation between two inner product spaces V and
W . Let T ∗ : W → V be the adjoint of T . Then we have the following orthogonal decompositions of
V and W ,
V = Ker T ⊕ Ran T ∗
and
W = Ker T ∗ ⊕ Ran T.
Equivalently this can be written as

(Ker T )⊥ = Ran T ∗ , (Ker T ∗ )⊥ = Ran T.

Before the proof let us notice that by the defintion of T ∗ we have

(T ∗ )∗ = T

i.e, T → T ∗ is like a mirroring, also the orthogonal complement V1⊥ of a linear subspace V1 ⊂ V has
the property
(V1⊥ )⊥ = V1
so W → W ⊥ is also like a mirroring. We thus need only to prove the first relation V = Ker T ⊕
Ran T ∗ .

Proof. The relation V = Ker T ⊕ Ran T ∗ is equivalent to Ker T = (Ran T ∗ )⊥ . We prove first

Ker T ⊆ (Ran T ∗ )⊥ .
5

Let v ∈ Ker T . Then T v = 0. We have to prove that v is orthogonal to the range Ran T ∗ . For any
T ∗ w ∈ Ran T ∗ , w ∈ W we compute
hv, T ∗ wi = hT v, wi = h0, wi = 0,
thus v ⊥ T ∗ w for any T ∗ w ∈ Ran T ∗ , and v ∈ (Ran T ∗ )⊥ proving the inclusion. Now we prove the
reverse inclusion
(Ran T ∗ )⊥ ⊆ Ker T.
Let v ∈ (Ran T ∗ )⊥ , i.e. v is orthogonal to Ran T ∗ . We prove that v ∈ Ker T , i.e. T v = 0. We
compute the paring of T v with arbitrary w ∈ W ,
hT v, wi = hv, T ∗ wi,
which is zero since v is orthogonal to T ∗ w ∈ Ran T ∗ W . So T v is orthogonal to any w ∈ W ,
therefore T v = 0, and v ∈ Ker T . This completes the proof.
Corollary 3. Let T : V → W be a linear operator between finite dimensional inner product vector
spaces. Then the rank of T and T ∗ are equal, Rank T = Ran T ∗ . In particular if V = W then we
have also dim Ker T = dim Ker T ∗ .
We follow the same proof and the notation as that of the Rank Theorem for linear transformations;
see Theorem 4 in Chapter 2 of the lecture notes. Actually the same proof along with the above
theorem that V = Ker T ⊕ T ∗ W will prove our present statement.

Proof. Let T1 = TT ∗ V be the restriction of T on the subspace T ∗ V ⊂ V . Thus T1 : T ∗ W → W , and


T1 : T ∗ W → T V since it is a restriction of T . We claim that T1 is injective. This is the same proof
as in the proof of Rank Theorem in Chapter 2 (see the lecture notes). Indeed suppose T1 v = 0 for
v ∈ T ∗ W ⊂ V . We have v = T ∗ w for some w ∈ V . Then
0 = T1 v = T v = T (T ∗ w) = T ∗ T w
and
0 = hw, T T ∗ wi = hT ∗ w, T ∗ wi = kT ∗ wk2
and
T ∗w = 0
i.e. v = T ∗ w = 0. Hence T1 : T ∗ W → T V is an injective map and
dim T V ≤ dim T ∗ W.
Replacing T ∗ by T we get also
dim T V ≤ dim T ∗ W.
This proves rank T = dim T V = dim T ∗ W = rank T ∗ .
If V = W then we have dim V = dim Ker T + rank T ∗ = dim W = dim Ker T ∗ + rank T , and
rank T = rank T ∗ as we just proved, thus dim Ker T = dim Ker T ∗ .
6

As a consequence we see that if A is a real m × n-matrix as linear transformation A : Rn → Rm


between the Euclidean spaces, then Rank A = Ran At since A∗ = At for real matrices. This is also
proved in Chapter 2 using row reductions.

Remark 6.1. Note that the theorem is true for both real and complex vector spaces. Also, with some
slight modification, (using the so-called closure of a subspace) the above theorem holds also for a
linear tranformation T : V → W between two infinite-dimensional Hilbert spaces.

Example 6.1. The above theorem can be illustrated by Figure 6.1 below.
7

T : V > W

Kert
✓ .

Rant

ker
TA W
'
-
i

T ✓

¥7
Figure 6.1: Orthogonality relation of Ker T and Ran T ∗
Index

adjoint (adjungerade transformationen), T ∗ , 3 Riesz Theorem, 2


Hermitian adjoint T ∗ , 3
linear functional (linjär funktional), 1 transpose (transponant), T t , 3

You might also like