Exercises: Similarity Transformation

Problem 1. Diagonalize the following matrix:

 
1 −1
A =
2 4

Solution. Matrix A has two eigenvalues λ1 = 3 and λ2 = 2. Since (i) A is a 2 × 2 matrix and
(ii) it has 2 distinct eigenvalues, we can apply the diagonalization method
 we discussed in class.
1
Specifically, we obtain an arbitrary eigenvector v1 of λ1 , say v1 = and, and an arbitrary
−2
 
1
eigenvector v2 of λ2 , say v2 = . Then, we form:
−1
 
1 1
Q =
−2 −1

by using v1 and v2 as the first and second columns, respectively. Q has the inverse:
 
−1 −1 −1
Q =
2 1

We thus obtain the following diagonalization of A:

A = Q diag[3, 2] Q−1 .

Problem 2. Consider again the matrix A in Problem 5. Calculate At for any integer t ≥ 1.

Solution. We already know that A:

A = Q diag[3, 2] Q−1 .

Hence:

At = Q diag[3t , 2t ] Q−1
  t  
1 1 3 0 −1 −1
=
−2 −1 0 2t 2 1
−3t + 2t+1 −3t + 2t
 
=
2 × 3t − 2t+1 2 × 3t − 2t

 
0 0 1
Problem 3. Diagonalize the matrix A = 0
 1 0 .
1 0 0

Solution. Recall that all symmetric matrices are diagonalizable. A is a 3 × 3 matrix. The key is
to find three linearly independent eigenvectors.

1
 From the solution of Problem
 1, we know that A has eigenvalues λ1 = 1 and λ2 = −1.
x1
EigenSpace(λ1 ) includes all  x2  satisfying
x3
x1 = u
x2 = v
x3 = u

 u, v ∈ R. The vector space EigenSpace(λ

for any  1 )has dimension 2 with a basis {v1 , v2 } where
1 0
v1 =  0  (given by u = 1, v = 0) and v2 =  1  (given by u = 0, v = 1).
1 0
 
x1
Similarly, EigenSpace(λ2 ) includes all  x2  satisfying
x3
x1 = u
x2 = 0
x3 = −u
for
 any  u ∈ R. The vector space EigenSpace(λ2 ) has dimension 1 with a basis {v3 } where v3 =
1
 0  (given by u = 1).
-1
So far, we have obtained three linearly independent eigenvectors v1 , v2 , v3 of A. We can then
apply the diagonalization method exemplified in Problem 5 to diagonalize A. Specifically, we form:
 
1 0 1
Q =  0 1 0 
1 0 −1
Q has the inverse:
 
1/2 0 1/2
Q−1 =  0 1 0 
1/2 0 −1/2
We thus obtain the following diagonalization of A:
A = Q diag[1, 1, −1] Q−1 .

Problem 4. Suppose that matrices A and B are similar to each other, namely, there exists P
such that A = P −1 BP . Prove: if x is an eigenvector of A under eigenvalue λ, then P x is an
eigenvector of B under eigenvalue λ.

Solution. By definition of similarity, we know A = P −1 BP . We proved in the lecture that λ

must also be an eigenvalue of B. Since x is an eigenvector of A under λ, we know:
Ax = λx ⇒
−1
P BP x = λx ⇒
B(P x) = λ(P x)

2
which completes the proof.

Problem 5. Suppose that an n × n matrix A has n linearly independent eigenvectors v 1 , v 2 , ..., v n .

Prove: for any n × 1 vector x, Ax is a linear combination of v 1 , v 2 , ..., v n .

Solution. Assume that v i (i ∈ [1, k]) is an eigenvector of A under eigenvalue λi . We have Av i =

λi v i . Since v 1 , v 2 , ..., v n are linearly independent, we know that x must be a linear combination
v 1 , v 2 , ..., v n . Namely, there exist c1 , ..., cn such that

x = c1 v 1 + c2 v 2 + ... + cn v n ⇒
Ax = c1 Av 1 + c2 Av 2 + ... + cn Av n ⇒
Ax = c1 λ1 v 1 + c2 λ2 v 2 + ... + cn λn v n .

which completes the proof.

Problem 6. Prove or disprove: if an n × n matrix A has rank n, then it must have n independent
eigenvectors.

Solution. False.  
1 1
Consider n = 2 and A = . It has only one distinct eigenvalue 1. Thus, any eigenvector
0 1
v of A must satisfy:

(A − I)x = 0 ⇒
 
0 1
x = 0
0 0
  
t
Thus, any eigenvector of A must have the form t ∈ R, t 6= 0 . This set of vectors has a
0
dimension of 1.
 
1 1 0
Problem 7. Prove that A =  0 1 0  is not diagonalizable.
0 0 2

Solution. A has two eigenvalues λ1 = 1 and λ2 = 2. Let v 1 be an eigenvector of λ1 . v 1 must

satisfy:

(A − λ1 I)v 1 = 0 ⇒
 
0 1 0
 0 0 0  v1 = 0 ⇒
0 0 1

Hence, the set of eigenvectors of λ1 is:

  
 t 
 0  t ∈ R, t 6= 0
0
 

This set has dimension 1.

3
 Let v 2 be an eigenvector of λ2 . v 2 must satisfy:

(A − λ1 I)v 2 = 0 ⇒
 
−1 1 0
 0 −1 0  v 2 = 0 ⇒
0 0 0

Hence, the set of eigenvectors of λ2 is:

  
 0 
 0  t ∈ R, t 6= 0
t
 

This set also has dimension 1.

It thus follows that the largest number of linearly independent eigenvectors of A is 1 + 1 = 2.
Therefore, A is not diagonalizable.

Problem 8. Let A, B, and C be three n × n matrices for some integer n. Prove that if A is
similar to B and B is similar to C, then A is similar to C.

Solution. From the fact that A is similar to B and B is similar to C, we know:

A = P −1 BP

and

B = Q−1 CQ.

Hence:

A = P −1 Q−1 BQP = (QP )−1 B(QP )

which completes the proof.

Problem 9. Decide whether

 
1 −1
A =
2 4

is similar to
 
3 1
B = .
0 2

Solution 1. From Problem 1, we know that A has distinct eigenvalues 3 and 2. Hence, A is
similar to the diagonal matrix diag[3, 2]. On the other hand, B clearly also has eigenvalues 3 and
2, and thus, is also similar to diag[3, 2]. From the result of Problem 8, we know that A is similar
to B.

4
  
x y
Solution 2. We will try to find an invertible matrix P = that makes A = P BP −1
z w
hold. This is equivalent to AP = P B. Hence:
     
1 −1 x y x y 3 1
= ⇒
2 4 z w z w 0 2
   
x−z y−w 3x x + 2y
=
2x + 4z 2y + 4w 3z z + 2w

This gives the following equation set:

x − z = 3x
y − w = x + 2y
2x + 4z = 3z
2y + 4w = z + 2w
    
x 
 −u/2 

 y   u/2 − v  
You can verify that the set of solutions 
  is   u ∈ R, v ∈ R .
z   
 u  

w v
 
 
−1 2
Let us try u = 2, v = 0. This gives P = . Since det(P ) 6= 0, we know that P is
2 0
invertible. We can now conclude that A is similar to B.