Algebra

Eigenvalues and Eigenvectors

FDA 2024-2025
Instructors: Nguyen Trung Thanh & Dam Tien Thanh (DS 66A-B)
Nguyen Tuan Long & Vu Huan (AI 66A-B)

Exercise

1. Let    
" # 1 3 3 0 0 1
1 −1
A= , B = 0 2 3 , C =  0 1 0 .
   
2 4
0 0 3 1 0 0

Compute A10 , B10 , C10 ?

Hint. Diagonalize the matrices and use the formulation: Am = PDm P−1 .

2. Prove that any two diagonal matrices with the same diagonal entries (possibly in a different
order) are similar to each other.
Hint. Suppose that A and B are two diagonal matrices with the same set of diagonal entries
λ1 , . . . , λn . Note that both the matrices are diagonalizable. We do steps diagonalizing matrix
B to attain the following form B = PDP−1 , in which the entries of D are ordered such that
D ≡ A. Consequently, B = PAP−1 .

3. 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 Px is an eigenvector
of B under eigenvalue λ.
Hint. We have B(Px) = PAP−1 Px = PAx = λ(Px).

4. Prove that if A is diagonalizable then AT , Am , m ∈ Z>0 , and A−1 are diagonalizable.

Hint. If A = PDP−1 with diagonal matrix D, then we have that A⊤ = (PDP−1 )⊤ = (P−1 )⊤ DP⊤ =
(P⊤ )−1 DP⊤ .

5. Prove that if A ∼ B then A2 ∼ B2 . Is the converse also true?

Solution. Counterexample. Consider
" # " #
0 0 0 1
A= , B=
0 0 0 0

6. Let A be an n×n square matrix. Prove: A and AT have exactly the same eigenvalues.
Solution: Recall that an eigenvalue of a matrix is a root of the matrix’s characteristic equa-
tion, which equates the matrix’s characteristic polynomial to 0. It suffices to show that the
characteristic polynomial of A is the same as that of AT . In other words, we want to show
that det(A − λI) = det(AT − λI). This is true because A − λI = (AT − λI)T .

7. Let A be an n × n square matrix. Prove: A−1 exists if and only if 0 is not an eigenvalue of A.
Solution:
If-Direction. The objective is to show that if 0 is not an eigenvalue of A, then A−1 exists, na-
mely, the rank of A is n. Suppose, on the contrary, that the rank of A is less than n. Consider
the linear system Ax = 0 where x is an n × 1 matrix. The hypothesis that rank(A) < n indi-
cates that the system has infinitely many solutions. In other words, there exists a non-zero
x satisfying Ax = 0x = 0. This, however, indicates that 0 is an eigenvalue of A, which is a

1
Instructors: Nguyen Trung Thanh & Dam Tien Thanh (DS 66A-B)
Nguyen Tuan Long & Vu Huan (AI 66A-B) FDA 2024-2025
contradiction.
Only-If Direction. The objective is to show that if A−1 exists, then 0 is not an eigenvalue of A.
The existence of A−1 means that the rank of A is n, which in turn indicates that Ax = 0 has
a unique solution x = 0. In other words, there is no non-zero x 0 satisfying Ax = 0x = 0,
namely, 0 is not an eigenvalue of A.

8. Let A be an n × n square matrix such that A−1 exists. Prove: if λ is an eigenvalue of A, then
1/λ is an eigenvalue of A−1 .
Solution: Since λ is an eigenvalue of A, there is a non-zero n × 1 matrix x satisfying

Ax = λx ⇒ A−1 Ax = λA−1 x ⇒ x = λA−1 x ⇒ A−1 Ax = (1/λ)x.

9. Prove: if A2 = I, then the eigenvalues of A must be 1 or −1.

Solution: Consider any eigenvalue λ of A, and let x be an arbitrary eigenvector of A corres-
ponding to λ. Hence, we have:

Ax = λx ⇒ A2 x = λAx ⇒ Ix = λAx ⇒ x = λAx.

Note that λAx = λ(λx) = λ2 x. Hence, we have x = λ2 x. As x ̸= 0, it follows that λ2 = 1.

10. Suppose that λ1 and λ2 are two distinct eigenvalues of matrix A. Furthermore, suppose
that x1 is an eigenvector of A under λ1 , and that x2 is an eigenvector of A under λ2 . Prove:
there does not exist any real number c such that cx1 = x2 .
Solution: Assume, on the contrary, that such a c exists. Since Ax1 = λ1 x1 , we have A(cx1 ) =
λ1 (cx1 ), which leads to Ax2 = λ1 x2 . On the other hand, Ax2 = λ2 x2 . Therefore, λ1 = λ2
(remember x2 cannot be 0), giving a contradiction.

11. Prove or disprove: if an n × n matrix A has rank n, then it must have n independent eigen-
vectors.
Solution: False. Consider the matrix
" #
1 1
A=
0 1

12. Suppose that λ1 and λ2 are two distinct eigenvalues of matrix A. Furthermore, suppose
that x1 is an eigenvector of A under λ1 , and that x2 is an eigenvector of A under λ2 . Prove:
x1 + x2 is not an eigenvector of A.
Solution: Assume, on the contrary, that x1 + x2 is an eigenvector under some eigenvalue
λ3 . This means that

A(x1 + x2 ) = λ3 (x1 + x2 ) ⇒ Ax1 + Ax2 = λ3 (x1 + x2 ) ⇒ λ1 x1 + λ2 x2 = λ3 (x1 + x2 ).

Hence, (λ1 − λ3 )x1 = (λ3 − λ2 )x2 . As λ1 ̸= λ2 , at least one of λ1 − λ3 and λ3 − λ2 is non-zero.

Without loss of generality, suppose λ3 − λ2 ̸= 0, which gives

λ1 − λ3
x1 = x2
λ3 − λ2
In a previous problem, we already showed that the above is impossible, thus giving a con-
tradiction.

2
 Instructors: Nguyen Trung Thanh & Dam Tien Thanh (DS 66A-B)
FDA 2024-2025 Nguyen Tuan Long & Vu Huan (AI 66A-B)
13. Let " #
a b
A=
c d

Show that

(a) A is diagonalizable if (a − d)2 + 4bc > 0

(b) A is not diagonalizable if (a − d)2 + 4bc < 0

Hint: Compute the characteristic polynomial of A, which is a quadratic polynomial. If the

discriminant of this quadratic polynomial is positive, that is (a − d)2 + 4bc > 0, then A has
two distinct eigenvalues and thus diagonalizable.

14. Prove that if A is a diagonalizable matrix, then the rank of A is the number of nonzero
eigenvalues of A.

15. Prove that similar matrices have the same rank.

16. Prove that if cA (λ) is the characteristic polynomial of a matrix A, then cA (A) = 0.
Hint: Suppose that cA (λ) = λn + an−1 λn−1 + · · · + a1 λ + a0 .. The Cayley Hamilton theorem
says that An + an−1 An−1 + · · · + a1 λA + a0 I = 0.. The proof uses the adjugate identity
det(tI − A) = (tI − A)adj(tI − A), and then just plugs in t = A.
17. Prove that if a, b, c, and d are integers such that a + b = c + d, then
" #
a b
A=
c d

has integer eigenvalues λ1 = a + b and λ2 = a − c.

Hint: Consider the characteristic polynomial of A: cA (λ) = (a−λ)(d−λ)−bc = λ2 −(a+d)λ+
ad−bc. The solutions of this polynomial are of the form: λ = 12 (a+d± (a + d)2 − 4(ad − bc)).
p

Since a+b = c+d, we have that (a+d)2 −4(ad−bc) = (b+c)2 , and thus λ = 12 (a+d±(b+c)).
Hence, λ is either a + b or a − c.

18. Consider the matrix  

1 0 0 0
a 1 0 0
A=
 
e b 2 0

f g c 2
Under which conditions on the unknowns is the matrix A diagonalizable?
Hint: One can see that the matrix A has two eigenvalues λ = 1, 2. A is diagonalizable if and
only if it has 4 basic eigenvectors.
We solve the linear system (A − λI)x = 0 to find the corresponding eigenvectors with
respect to λ. For λ = 1, we formulate the augmented matrix as follows.
 
0 0 0 0 0  
a 0 0 0 0
a 0 0 0 0

 →  e b 1 0 0
  
e b 1 0 0

f g c 1 0
f g c 1 0

Hence, the linear system can be written as



 ax1 =0
ex1 + bx2 + x3 =0

 fx + gx + cx + x = 0
1 2 3 4

3
Instructors: Nguyen Trung Thanh & Dam Tien Thanh (DS 66A-B)
Nguyen Tuan Long & Vu Huan (AI 66A-B) FDA 2024-2025
If a ̸= 0 then x1 = 0. Let x2 = t, t ∈ R, then x3 = −bt and x4 = (−g + bc)t. Hence the system
has only one basic feasible solution. If a = 0, the linear system becomes:

ex1 + bx2 + x3 =0
fx1 + gx2 + cx3 + x4 = 0

One needs two parameters to express the solutions of the systems: x1 = t, x2 = s, t, s ∈ R.

This means that the number of basic solutions is 2.
Similarly, for λ = 2, we formulate the augmented matrix as follows.
       
−1 0 0 0 0 −1 0 0 0 0 −1 0 0 0 0 −1 0 0 0 0
 a −1 0 0 0  0 −1 0 0 0  0 −1 0 0 0  0 −1 0 0 0
→ → →
       
 e b 0 0 0 0 b 0 0 0 0 0 0 0 0 0 0 c 0 0
 
f g c 0 0 0 g c 0 0 0 0 c 0 0 0 0 0 0 0

If c = 0, the number of leading ones of the matrix is 2 and thus the system has 2 basic
solutions.
Therefore, A is diagonalizable if and only if a = c = 0.

19. A three by three matrix B is known to have eigenvalues 0, 1 and 2. Give answer to the
following question.
a) The rank of B
b) The determinant of BT B
c) The eigenvalues of BT B
d) The eigenvalues of (B2 + I)−1
Solution:
a) B has 0 as an eigenvalue and is therefore singular (not invertible). Since B is a three by
three matrix, this means that its rank can be at most 2. Since B has two distinct nonzero
eigenvalues, its rank is exactly 2.
b) det(B) = 0
c) There is not enough information to find the eigenvalues of BT B
d) If p(t) is a polynomial and if x is an eigenvector of A with eigenvalue λ, then p(A)x = p(λ)x.
We also know that if λ is an eigenvalue of A then 1/λ is an eigenvalue of A−1 . Hence the
eigenvalues of (B2 + I)−1 are 1, 1/2 and 1/5.

20. Let a, b ∈ R and  

a b b
A = b a b
 
b b a
Compute An for all n ⩾ 1.
Solution: The case of b = 0 is easy because the matrix is diagonal. We assume now that
b ̸= 0.
det(A − λI) = (a − λ)3 + 2b3 − 3(a − λ)b2 = (a − b − λ)2 (a + 2b − λ)

Hence, the matrix A has two eigenvalues λ1 = a − b and λ2 = a + 2b.

Consider λ1 = a − b.
(A − (a − b)I)x = 0 ⇐⇒ x1 + x2 + x3 = 0.

4
 Instructors: Nguyen Trung Thanh & Dam Tien Thanh (DS 66A-B)
FDA 2024-2025 Nguyen Tuan Long & Vu Huan (AI 66A-B)
Thus, every solution x of the system is of the form: x1 = −t − s, x2 = t, x3 = s, t, s ∈ R, or
   
−1 −1
x = t 1  + s 0 
   
0 1

There are two basic eigenvectors v1 = (−1, 1, 0) and v2 = (−1, 0, 1).

Consider λ2 = a + 2b.

 
 −2x1 + x2 + x3 = 0 x1 − 2x2 + x3 = 0
(A − (a + 2b)I)x = 0 ⇐⇒ x1 − 2x2 + x3 = 0 ⇐⇒

 x + x − 2x x2 − x3 =0
1 2 3 =0

Thus, every solution x of the system is of the form: x1 = t, x2 = t, x3 = t, t ∈ R. There is one

basic eigenvector v3 = (1, 1, 1).
Finally, we have
 n    −1
a b b −1 −1 1 (a − b)n 0 0 −1 −1 1
An =  b a b  =  1 0 1  0 (a − b)n 0  1 0 1
     
b b a 0 1 1 0 0 (a + 2b)n 0 1 1

21. Let    
0 1 1 1 −4 −4
A = −1 2 0 and B = 0 3 0
   
1 0 0 0 0 1

a) Compute A−1 .
b) Compute det(2A2 B), det(4A + B), det(2(A3 B−2 )).
Solution:  
0 0 2
1
A−1 = 0 1 1

2
2 −1 −1
1
det(A) = −2, det(B) = 3, det(2A2 B) = 23 .det(A)2 .det(B), det(2(A3 B−2 )) = 23 .det(A)3 . det(B) 2.

22. Given that the characteristic polynomial of a 3 × 3 matrix A is λ3 + 2λ2 + 4λ + 8. Compute

A4 ?
Solution: Applying Cayley-Hamilton theorem, we have that A3 + 2A2 + 4A + 8I = 0. Mul-
tiplying by A − 2I then gives A4 − 16I = 0, or A4 = 16I.

23. Assume the 2 × 2 matrix A is similar to an upper triangular matrix. If trace(A) = 0 =

trace(A2 ), show that A2 = 0.
Solution: We have that A = PDP−1 , A = PD2 P−1 , where
" # " #
a b 2a2 b(a + c)
D= , D =
0 c 0 c2

Since A2 and D2 are similar, and trace(A2 ) = 0, it holds that a2 + c2 = 0, or a = c = 0.

Hence D2 = 0 and thus A2 = 0.

5
Instructors: Nguyen Trung Thanh & Dam Tien Thanh (DS 66A-B)
Nguyen Tuan Long & Vu Huan (AI 66A-B) FDA 2024-2025

True-False Questions

1. If A and B are similar invertible matrices, then A−1 and B−1 are similar. (T)

2. If A is diagonalizable, then there is a unique matrix P such that P−1 AP is diagonal. (F)

3. If A is invertible then A is diagonalizable (F)

4. If A is diagonalizable and invertible, then A−1 is diagonalizable. (T)

5. If λ is an eigenvalue of a matrix A, then the linear system (A − λ)x = 0 has only the trivial
solution. (F)

6. If the characteristic polynomial of a matrix A is λ2 + 1, then A is invertible. (T)

7. If 0 is an eigenvalue of a matrix A, then A2 is singular (A singular matrix refers to a matrix

whose determinant is zero). (T)

8. The eigenvalues of a matrix A are the same as the eigenvalues of the reduced row echelon
form of A. (F)

9. If 0 is an eigenvalue of a matrix A, then the set of columns of A is linearly independent. (F)

10. If λ is an eigenvalue of A, and s is a scalar, then λ − s is an eigenvalue of A − sI. (T)

11. If λ is an eigenvalue of A, and s is a scalar, then sλ is an eigenvalue of sA. (T)

12. Every matrix A ∈ Rn×n , where n is odd, always has at least one eigenvalue. (T)

13. If λ is an eigenvalue of A, then λr is an eigenvalue of Ar , for r ∈ Z>0 . (T)

14. If Ar = 0 for some integer r ∈ Z>0 , then all of A’s eigenvalues are 0. (T)

15. If A is invertible, then AB is similar to BA for all B. (T)

16. If A ∼ B and A2 = A then B2 = B. (T)

17. If A ∼ B and A is invertible then so is B. (T)

18. Let A denote an upper triangular matrix. Then

a) If all the main diagonal entries of A are distinct, then A is diagonalizable. (T)
b) If all the main diagonal entries of A are equal, then A is diagonalizable only if it is already
diagonal. (T)

19. Two diagonalizable matrices are similar if and only if they have the same eigenvalues. (T)

20. Let A be n × n with n distinct real eigenvalues. If AC = CA, then C is diagonalizable. (T)