24

5. Solutions to Exercise 5

Exercise 5.1. Suppose V is finite-dimensional. Then each operator on V has at most dim V
distinct eigenvalues.

Solution 5.1. Let T ∈ L(V ). Suppose λ1 , . . . , λm are distinct eigenvalues of T . Let v1 , . . . , vm

be corresponding eigenvectors. Then {v1 , . . . , vm } is linearly independent. Thus m ≤ dim V .

Exercise 5.2.

(1) Suppose S, T ∈ L(V ) are such that ST = T S. Prove that Nul(S) is invariant under T .

Solution 5.2. For any v ∈ Nul(S), S(v) = 0. Since ST = T S, S(T (v)) = T (S(v)) =
T (0) = 0. Then T (v) ∈ Nul(S). Then Nul(S) is invariant under T .
(2) Suppose S, T ∈ L(V ) are such that ST = T S. Prove that im(S) is invariant under T .

Solution 5.2. For any v ∈ im(S), there exists w ∈ V such that S(w) = v. Since ST = T S,
T (v) = T (S(w)) = S(T (w)). Then T (v) ∈ im(S). Then im(S) is invariant under T .

Exercise 5.3. See the proof of Theorem 5.4.1. Let v ∈ V . Let {[v1 ], . . . , [vk ]} be a basis
V / Span(v). Please show that {v, v1 , . . . , vk } is linearly independent using the definition of
linearly independence.

Solution 5.3. Set up the equation in V :

cv + c1 v1 + . . . + ck vk = 0. (**)

Then c1 v1 + . . . + ck vk = −cv ∈ Span(v). Then [c1 v1 + . . . + ck vk ] = 0 ∈ V / Span(v). So

c1 [v1 ] + . . . + ck [vk ] = [c1 v1 + . . . + ck vk ] = 0.

Since {[v1 ], . . . , [vk ]} is a basis of V / Span(v), c1 = . . . = ck = 0. Then in Equation (**), we

have cv = 0. Therefore c = 0. Then {v, v1 , . . . , vk } is linearly independent. □

Exercise 5.4. Suppose V is a finite-dimensional complex vector space and T ∈ L(V ). Prove
that T has an invariant subspace of dimension k for each k = 1, . . . , dim V .
 25
[ ]
Proof. Let n = dim V . By Theorem 5.4.1, there is a basis B = {v1 , . . . , vn } such that T
B←B
is upper-triangular. Then
T (v1 ) ∈ Span(v1 ),

T (v2 ) ∈ Span(v1 , v2 ),

......

T (vn ) ∈ Span(v1 , . . . , vn ).

Then Wk = Span(v1 , . . . , vk ) is the desired T -invariant subspace of dimension k for any k =

1, . . . , n. □

Exercise 5.5. Suppose W is a complex vector space and T ∈ L(W ) has no eigenvalues. Prove
that every subspace of W invariant under T is either {0} or infinite-dimensional.

Proof. Use contradiction. Assume that there is a finite-dimensional non-zero subspace U of W

invariant under T . Then by Theorem 5.4.2, T |U has an eigenvector in U . Then there exists
w ∈ U , λ ∈ C such that w ̸= 0 and T (w) = λw. Since w ∈ U ⊂ W , w is an eigenvector of T in
W . This is a contradiction. So the assumption is wrong. Then every subspace of W invariant
under T is either {0} or infinite-dimensional. □

Exercise 5.6. Let T ∈ L(C3 ) which is defined by the matrix

 
2 −2 0
 
 
A = 0 3 0.
 
1 6 2

Find a basis C3 to write T as an upper-triangular matrix.

      

 1 0 0 


      

     
Solution 5.6. Let V = C3 . Let S = e1 = 0, e2 = 1, e3 = 0 be the standard basis.

      


 

0 0 1
Then  
2 −2 0
[ ]  
 
T = 0 3 0.
S←S  
1 6 2
26

(1) Find an eigenvector of T on V : det(A − λI) = 0. Then (2 − λ)2 (3 − λ) = 0. So 2 and 3 are

eigenvalues. We use 2 as our current eigenvalue here. Solve the equation (A − 2I)X = 0
for X ∈ C3 .  
2−2 −2 0
 
 
 0 3−2 0 X = 0.
 
1 6 2−2
   
0 0
   
   
Solutions are X ∈ Span 0. We take v1 = 0 as our first eigenvector.
   
1 1
{[ ] [ ]}
(2) Consider the quotient space V / Span(v1 ). Consider the basis e1 , e2 where

   
1 0
[ ]   [ ]  
   
e1 = 0 + Span(v1 ), e2 = 1 + Span(v1 ).
   
0 0
   
[ ] 1 [ ] 0
V / Span(v1 ) ≃ C2 by e1 7→  , e2 7→  .
0 1
(3) Since
 
2
([ ]) [ ]   [ ]
 
T e1 = T (e1 ) = 0 = 2 e1 ,
 
1
 
−2
([ ]) [ ]   [ ] [ ]
 
T e2 = T (e2 ) =  3  = −2 e1 + 3 e2 .
 
6
 
{[ ] [ ]} 2 −2
Therefore under the basis e1 , e2 , the matrix of T is B =  .
0 3
(4) Find an eigenvector of T on V / Span(v1 ): det(B − λI) = 0. Then (2 − λ)(3 − λ) = 0.
So 2 and 3 are eigenvalues. We use 2 as our current eigenvalue here. Solve the equation
 27

(B − 2I)X = 0 for X ∈ C2 ≃ V / Span(v1 ).

 
2 − 2 −2
 X = 0.
0 3−2
   
1 1
Solutions are X ∈ Span  . We take v2 =  . By the isomorphism between C2 and
0 0
 
  1
1  
 
V / Span(v1 ),   is corresponding to 0 + Span(v1 ). We take a representative from the
0  
0
 
1
 
 
class to be v2 . For example we can pick v2 = 0.
 
0
{[ ]}
(5) Consider the quotient space V / Span(v1 , v2 ). Consider the basis e2 where

 
0
[ ]  
 
e2 = 1 + Span(v1 , v2 ).
 
0
[ ]
V / Span(v1 , v2 ) ≃ C by e2 7→ 1.
(6) Since


−2
([ ]) [ ]   [ ]
 
T e2 = T (e2 ) =  3  = 3 e2 .
 
6
28
[ ]
Then e2 is an eigenvector of T in V / Span(v1 , v2 ). The vector is corresponding to the
 
0
 
 
class 1 + Span(v1 , v2 ). We take a representative from the class to be v3 . For example we
 
0
 
0
 
 
can pick v3 = 1.
 
0
      

 0 1 0 

      
     
(7) We now have a basis B = v1 = 0, v2 = 0, v3 = 1 . The change-of-basis PS←B =

      

 

1 0 0
 
0 1 0
 
 
0 0 1. Then the matrix of T is
 
1 0 0

 
2 1 6
[ ] [ ]  
−1  
T = PS←B T PS←B = 0 2 −2.
B←B S←S  
0 0 3