Answers/Solutions of Exercise 6 (Q1-8) (Version: October 31, 2014)

1. (a) The characteristic equation is (λ + 1)(λ − 3) = 0; eigenvalues are −1 and

3; {(0, 1)T } is a basis for E−1 and {(1, 2)T } is a basis for E3 .
(b) The characteristic equation is (λ − 2)2 = 0; the eigenvalue is 2; {(1, 1)T }
is a basis for E2 .
(c) The characteristic equation is λ2 − 4 = 0; eigenvalues are −2 and 2;
{(−2, 1)T } is a basis for E−2 and {(2, 1)T } is a basis for E2 .
(d) The characteristic equation is λ2 = 0; the eigenvalue is 0; {(1, 0), (0, 1)T }
is a basis for E0 .
(e) The characteristic equation is λ(λ − 2)2 = 0; eigenvalues are 0 and 2;
{(−1, 1, 0)T } is a basis for E0 and {(1, 1, 0)T } is a basis for E2 .
(f) The characteristic equation is (λ − 2)(λ2 − 9) = 0; eigenvalues are 2, −3
and 3; {(0, 0, 1)T } is a basis for E2 , {(−1, 3, 0)T } is a basis for E−3 and
{(1, 3, 0)T } is a basis for E3 .
(g) The characteristic equation is (λ − 1)3 = 0; the eigenvalue is 1; {(0, 0, 1)T }
is a basis for E1 .
(h) The characteristic equation is (λ + 1)(λ − 1)2 = 0; eigenvalues are −1 and
1; {(−1, −1, 1)T } is a basis for E−1 and {(1, 2, 0)T , (1, 0, 2)T } is a basis for
E1 .
(i) The characteristic equation is (λ − 1)(λ − 2)(λ − 3)(λ − 4) = 0; eigenvalues
are 1,2,3 and 4; {(0, 0, 0, 1)T }is a bais for E1 , {(0, 0, 1, 1)T } is a basis for
E2 , {(0, 2, 4, 3)T } is a basis for E3 and {(3, 9, 12, 8)T } is a basis for E4 .
(j) The characteristic equation is λ4 − 2λ2 + 1 = 0; eigenvalues are −1 and 1;
{(−1, 0, 1, 0)T , (0, −1, 0, 1)T } is a basis for E−1 and {(1, 0, 1, 0)T , (0, 1, 0, 1)T }
is a basis for E1 .
 
λ − a −b 

2. (a) det(λI − A) =   = λ2 + (−a − d)λ + (ad − bc)
−c λ − d
Hence m = −a − d = −tr(A) and n = det(A).
(b) Direct verification shows that A2 + mA + nI = 0.

3. (a) Let x be an eigenvector of A associated with λ, i.e. Ax = λx. We prove

that An x = λn x by induction on n.
It is given that A1 x = λ1 x. Assume that Ak x = λk x. Then

Ak+1 x = A(Ak x) = A(λk x) = λk Ax = λk λx = λk+1 x.

1
 By mathematical induction, An x = λn x and hence λn is an eigenvalue of
A for all positive integer n.

(b) Let x be an eigenvector of A associated with λ. Then

Ax = λx ⇒ x = A−1 (λx) = λA−1 x ⇒ 1

λ
x = A−1 x.

Thus 1
λ
is an eigenvalue of A−1 .

(c) λ is an eigenvalue of A ⇒ det(λI − A) = 0

⇒ det ((λI − A)T ) = 0
⇒ det(λI − AT ) = 0
⇒ λ is an eigenvalue of AT .

4. (a) Let x be an eigenvector of A associated with λ, i.e. Ax = λx and x is a

nonzero vector. Then

A2 = A ⇒ A2 x = Ax ⇒ λ2 x = λx ⇒ λ(λ − 1)x = 0

Since x is nonzero, λ = 0 or 1.
( )
a b
(b) Since A has 2 distinct eigenvalues, it is diagonalizable. Let P =
c d
( )
1 0
be an invertible matrix such that P −1 AP = . Then
0 0
( )( )( )−1 ( )
a b 1 0 a b 1 ad −ab
A= = where ad−bc ̸= 0.
c d 0 0 c d ad − bc cd −cb
( )
r s
We can simplify the expression to A = where st = r(1 − r).
t 1−r

5. (a) Let x be a nonzero eigenvector of A associated with λ, i.e. Ax = λx.

A2 = 0 ⇒ A2 x = 0x ⇒ A(λx) = 0 ⇒ λ2 x = 0
Since x is nonzero, λ = 0.
(b) No. Suppose A is diagonalizable. Then there exists invertible P such that
P −1 AP = 0. Then A = P 0P −1 = 0, a contradiction.
(c) Consider the vector equation

au + bAu = 0. (∗)

Pre-multiplying A to both side of (∗), we have

A(au + Au) = A0 ⇒ aAu = 0. (∵ A2 = 0.)

2
 As Au ̸= 0, a = 0. Substituting a = 0 into (∗), we have bAu = 0 and
hence b = 0. Since (∗) has only the trivial solution, u and Au are linearly
independent.
( )
(d) Let P = u Au . By (c), P is invertible. Since
( ) ( )
AP = Au A2 u = Au 0

and ( )
0 0 ( ) ( )
P = 0u + Au 0u + 0Au = Au 0 ,
1 0
( ) ( )
0 0 −1 0 0
AP = P which implies P AP = .
1 0 1 0

6. (a) Since det(−I − A) = 0, −1 is an eigenvalue of A.

(b) {(1, 1, 0)T , (0, 0, 1)T } is a basis for E−1 and hence dim(E−1 ) = 2.
 
1 0 0
(c) For example, B = 0  1 0.
0 0 3

7. (a) Since det(2I − A) = 0, 2 is an eigenvalue of A.

(b) {(1, 2, 0)T , (−3, 0, 1)T } is a basis for the eigenspace associated with 2.
(c) Let E2 be the eigenspace of A associated with 2 and let Eλ′ be the eigenspace
of B associated with λ.
Since E2 and Eλ′ are subspaces of R3 and have dimension 2, they are two
planes in R3 that contain the origin. So E2 ∩ Eλ′ is either a line through
the origin or a plane containing the origin. In both cases, we can find a
nonzero vector u ∈ E2 ∩ Eλ′ , i.e. Au = 2u and Bu = λu, such that

(A + B)u = Au + Bu = 2u + λu = (2 + λ)u.

So 2 + λ is an eigenvalue of A + B.

8. Note that for i = 1, 2, . . . , n, An ui = An−1 ui+1 = · · · = Ai un = 0.

Let v ∈ Rn be an eigenvector of A associated with eigenvalue λ, i.e. Av = λv.
Since {u1 , u2 , . . . , un } is a basis for Rn ,

v = c1 u1 + c2 u2 + · · · + cn un

for some c1 , c2 , . . . , cn ∈ R. Then

An v = c1 An u1 + c2 An u2 + · · · + cn An un = 0.

3
From the proof of Question 6.3(a), An v = λn v. Since v ̸= 0, λ = 0. Hence we
have shown that A has only one eigenvalue 0.
As λ = 0, we get Av = 0. Then

0 = Av = c1 Au1 + c2 Au2 + · · · + cn Aun = c1 u2 + c2 u3 + · · · + cn−1 un .

Since u2 , u3 , . . . , un are linearly independent, c1 = 0, c2 = 0, . . . , cn−1 = 0, i.e.

v = cn un . Hence all eigenvectors of A are scalar multiples of un .