22

Solutions to Exercises 7

Exercise 7.1. Consider Cn with the dot product. Define T ∈ Cn by

   
 z1   0 
   
 z2   z1 
   
T  .  =  . .
 ..   .. 
   
   
zn zn−1

 
z
 1 
 . 
Find a formula for T ∗  .. . You cannot directly use Theorem 7.1.6.
 
zn

Solution 7.1. Choose the standard basis E = {e1 , . . . , en }. Then E is an orthonormal basis.
From the definition of T , we have that T (ei ) = ei+1 for i = 1, . . . , n − 1 and T (en ) = 0.
Let T ∗ (ej ) = a1j e1 +. . .+anj en . To find aij , we need to compute ⟨ei , a1j e1 + . . . + anj en ⟩ = aij .
Then

aij = ⟨ei , a1j e1 + . . . + anj en ⟩ = ⟨ei , T ∗ (ej )⟩ = ⟨T (ei ), ej ⟩




⟨ei+1 , ej ⟩ i = 1, . . . , n − 1,
=


⟨0, ej ⟩ i = n,



δi+1,j i = 1, . . . , n − 1,
=


0 i = n,

Then this means that aj−1,j = 1 for j = 2, . . . , n and aij = 0 otherwise. So T ∗ (e1 ) = 0 and
T ∗ (ej ) = ej−1 for j = 2, . . . , n. Then
    
 z1   z2 
    . 
 z2   .. 
    
T ∗  .  =  .
 ..  z 
    n 
    
zn 0
 23

Exercise 7.2. Suppose V is a complex inner product space and T ∈ L(V ) is a normal operator
such that T 9 = T 8 . Prove that T is self-adjoint and T 2 = T .

Solution 7.2. Since T 9 = T 8 , x9 − x8 is a multiple of the minimal polynomial of T . Then the

eigenvalues of T should be 0 or 1 or both. Then all eigenvalues of T are real. Since T is normal,
by Corollary 7.2.4 T is self-adjoint.
Since T is diagonal, the biggest Jordan block is of size 1. Then the minimal polynomial of T
should contain at most one factor x and one factor (x − 1). So the minimal polynomial divides
x(x − 1). Then T (T − I) = 0. Therefore T 2 = T .

Exercise 7.3. Let S, T ∈ L(V ) be self-adjoint. Show that ST is self-adjoint if and only if
ST = T S.

Solution 7.3.

(⇒): Since S, T , ST are self-adjoint, then S ∗ = S, T ∗ = T and (ST )∗ = ST . Then ST =

(ST )∗ = T ∗ S ∗ = T S.
(⇐): Since S ∗ = S, T ∗ = T and ST = T S, we have (ST )∗ = T ∗ S ∗ = T S = ST . So ST is
self-adjoint.

Exercise 7.4. Suppose V is a complex inner product space with V ̸= {0}. Show that the set
of self-adjoint operators on V is not a subspace of L(V ).

Solution 7.4. λ ̸= λ for most λ ∈ C. Therefore if T is self-adjoint, then (λT )∗ = λT ∗ = λT ̸=

λT in general. So self-adjoint operators don’t form a subspace of L(V ).

Exercise 7.5. Give an example of an operator T on a complex vector space such that T 9 = T 8
but T 2 ̸= T .

Solution 7.5. From the previous exercise, the biggest difference here is that T doesn’t have
to be diagonalizable. So to find a counter example we should look at those non-diagonalizable
24

operators. For example, let T ∈ C4 be defined by

 
0 1 0 0
 
0 0 1 0
 
A= .
0 0
 0 0 
 
0 0 0 1
   
0 0 0 0 0 0 1 0
   
0 0 0 0 0 0 0 0
   
A =A =
9 8
, and A = 
2
 ̸= A.
0 0 0 0
 0 0   0 0 
   
0 0 0 1 0 0 0 1

Exercise 7.6. Let V be a finite-dimensional complex vector space. Suppose that T is a normal
operator on V and that 3 and 4 are eigenvalues of T . Prove that there exists a vector v ∈ V
√
such that ∥v∥ = 2 and ∥T v∥ = 5.

Solution 7.6. Since T is normal, there exists an orthonormal basis such that all basis vectors
are eigenvectors. Let v1 be the one with eigenvalue 3 and v2 be the one with eigenvalue 4. Let
v = v1 + v2 . Then
√ √ √
∥v∥ = ∥v∥ = ∥v1 ∥2 + ∥v2 ∥2 = 2,
2

and
√
∥T v∥ = ∥T (v1 + v2 )∥ = ∥3v1 + 4v2 ∥ = ∥3v1 ∥2 + ∥4v2 ∥2 = 5.

Exercise 7.7. Give an example of an operator T ∈ L(C4 ) such that T is normal but not
self-adjoint.
 25

Solution 7.7. There are plenty based on Proposition

 7.2.2. Here
 is one: Let E be an or-

i 
 
[ ]  
 0 
thonormal basis. T is defined by T = A = 

. It is not self-adjoint since

E←E  0 
 
 
0
 
−i 
 
 
 0 
AH = 

 ̸= A. However AAH = AH A.

 0 
 
 
0

Exercise 7.8. Let T ∈ L(V, W ). Please prove that

(1) T is injective if and only if T ∗ is surjective.

(2) T is surjective if and only if T ∗ is injective.

Solution 7.8.

(1) (⇒): Assume T is injective. Assume that T ∗ is not surjective. Then im(T ∗ ) ⊊ V . Choose
a basis of im(T ∗ ), extend it to be a basis of V , and apply the Gram-Schmidt procedure to
get an orthonormal basis. Pick v to be any one basis vector outside im(T ∗ ). Since the first
dim im(T ∗ ) basis vectors form an orthonormal basis of im(T ∗ ), v should be orthogonal to
all vectors in im(T ∗ ). Then for any w ∈ W , ⟨T v, w⟩ = ⟨v, T ∗ w⟩ = 0. So T v = 0. Then by
T being injective, v = 0. This is a contradiction. So im(T ∗ ) = V . Then T ∗ is surjective.
(⇐): Assume T ∗ is surjective. Assume T v = 0 for some v ∈ V . Then for any w ∈ W ,
⟨v, T ∗ w⟩ = ⟨T v, w⟩ = 0. Since T ∗ is surjective, this means that ⟨v, u⟩ = 0 for any u ∈ V .
Then v = 0. So T is injective.
(2) It follows from the first part by T = (T ∗ )∗ .
26

Exercise 7.9. Consider C3 with the dot product. Let E be the standard basis. Let T ∈ L(C3 )
be defined by
 
1 1 0
[ ]  
 
T = 0 1 1.
E←E  
1 0 1
Please find an orthonormal basis such that the matrix of T is diagonal, or prove that such a
basis doesn’t exist.

Solution 7.9. Since E is an orthonormal basis,

 
1 0 1
[ ] [ ]H  
 
T∗ = T = 1 1 0.
E←E E←E  
0 1 1

Since        
1 1 0 1 0 1 2 1 1 1 0 1 1 1 0
       
       
0 1 11 1 0 = 1 2 1 = 1 1 00 1 1,
       
1 0 1 0 1 1 1 1 2 0 1 1 1 0 1
T is normal. Then T can be diagonalized by an orthonormal basis, which are also eigenvectors.
Solve
([ ] )
det T − λI = 0.
E←E
√ √
1+ 3i 1− 3i
Then λ = 2, 2
and 2
. These are three eigenvalues.

   
1−2 1 0 1
   
   
λ = 2: Solve  0 1−2 1 X = 0. The solution is X ∈ Span 1. Then we choose
   
1 0 1−2 1
 √ 
1/ 3
 √ 
 
v1 = 1/ 3 since we need ∥v1 ∥ = 1.
 
√
1/ 3
 27
 √   √ 
−1+ 3i
1− 2
1+ 3i
1 0
√    2√ 
 √   
λ = 2 : Solve 
1+ 3i
0 1− 2
1+ 3i
1 X = 0. The solution is X ∈ Span  −1−2 3i .
 √
  
1 0 1− 2
1+ 3i
1
 √ 
−1+
√ 3i
 2 √3 
 
Then we choose v2 =  −1−√ 3i  since we need ∥v2 ∥ = 1.
 2 3 
√1
 √
3
  √ 
−1− 3i
1− 1− 3i
1 0
√  2   2√ 
 √   
λ = 1−2 3i : Solve  0 1− 1− 3i
1 X = 0. The solution is X ∈ Span  −1+2 3i .
 2
√
  
1 0 1− 1− 3i
2
1
 √ 
−1−
√ 3i
 2 √3 
 −1+ 3i 
Then we choose v2 =  √  since we need ∥v3 ∥ = 1.
 2 3 
√1
3
Then B = {v1 , v2 , v3 } is the orthonormal basis which make T diagonal:
 

[ ] 2 
 √ 
T =

1+ 3i .

B←B  2
√

1− 3i
2