Homework exercise for Chapter 4

Linear Algebra
November 8, 2024

Section 4.1
Question 17
(a)
1. To compute the kernel of L, we need to find for which vector x ∈ R3 , L(x) = 0.
2. The linear transformation in (a) is L(x) = (x3 , x2 , x1 )T . The equation L(x) = (x3 , x2 , x1 )T = 0 gives

x 3 = 0

x2 = 0

x1 = 0


The only vector mapped to 0 by the linear transformation L is 0. Thus, ker(L) = {0}.
3. To compute the range of L, L(R3 ), first observe the linear transformation L:

L(x) = (x3 , x2 , x1 )T

We can produce any vector y ∈ R3 since x3 , x2 , x1 can be set to arbitrary real numbers. So, we guess
the range of L is R3 .
4. For any vector x ∈ R3 , L(x) = (x3 , x2 , x1 )T ∈ R3 . So, we can conclude that L(R3 ) ⊆ R3 .
5. For any vector y ∈ R3 , can we find a vector x ∈ R3 , such that L(x) = y? The answer is yes. We can
always find x = (y3 , y2 , y1 )T such that L(x) = y. Thus, we can conclude that L(R3 ) ⊇ R3
6. By previous discussion, we can conclude that the range of L, L(R3 ), is R3 .

(b)
1. To compute the kernel of L, we need to find for which vector x ∈ R3 , L(x) = 0.
2. The linear transformation in (a) is L(x) = (x1 , x2 , 0)T . The equation L(x) = (x1 , x2 , 0)T = 0 gives
(
x1 = 0
x2 = 0

x3 can be any real number. So, the vectors mapped to 0 by the linear transformation L are in the
form (0, 0, α)T . Thus, ker(L) = {(0, 0, α)T | ∀α ∈ R}.
3. To compute the range of L, L(R3 ), let’s observe the linear transformation L first.

L(x) = (x1 , x2 , 0)T

We can produce any vector in the form (α, β, 0)T for any α, β ∈ R. So, we guess L(x) = {(α, β, 0)T |
∀α, β ∈ R}.

1
 4. For any vector y ∈ L(x), y can be expressed as (x1 , x2 , 0)T for some x1 , x2 ∈ R. So, y ∈ {(α, β, 0)T |
∀α, β ∈ R}. Thus, L(R3 ) ⊆ {(α, β, 0)T | ∀α, β ∈ R}
5. For any vector y ∈ {(α, β, 0)T | ∀α, β ∈ R}, let y = (α, β, 0)T , we can find the x = (α, β, 1.234567)
such that L(x) = y. So, y ∈ L(x). Thus, L(R3 ) ⊇ {(α, β, 0)T | ∀α, β ∈ R}

6. By previous discussion, we can conclude that the range of L, L(R3 ), is {(α, β, 0)T | ∀α, β ∈ R}.

(c)
1. To compute the kernel of L, we need to find for which vector x ∈ R3 , L(x) = 0.
2. The linear transformation in (a) is L(x) = (x1 , x1 , x1 )T . The equation L(x) = (x1 , x1 , x1 )T = 0 gives

x1 = 0

x1 = 0

x1 = 0


x2 , x3 can be any real number. So, the vectors mapped to 0 by the linear transformation L are in the
form (0, α, β)T . Thus, ker(L) = {(0, α, β)T | ∀α, β ∈ R}.
3. To compute the range of L, L(R3 ), let’s observe the linear transformation L first.

L(x) = (x1 , x1 , x1 )T

We can produce any vector in the form (α, α, α) for any α ∈ R. So, we guess L(x) = {(α, α, α)T | ∀α ∈
R}.
4. For any vector y ∈ L(x), y can be expressed as (x1 , x1 , x1 )T for some x1 ∈ R. So, y ∈ {(α, α, 0)T |
∀α ∈ R}. Thus, L(R3 ) ⊆ {(α, α, α)T | ∀α ∈ R}
5. For any vector y ∈ {(α, α, α)T | ∀α ∈ R}, let y = (α, α, α)T , we can find the x = (α, 0.765432, 1.234567)
such that L(x) = y. So, y ∈ L(x). Thus, L(R3 ) ⊇ {(α, α, α)T | ∀α ∈ R}
6. By previous discussion, we can conclude that the range of L, L(R3 ), is {(α, α, α)T | ∀α ∈ R}.

Question 21
Proof of only-if part (⇒)
1. We need a important property of vector space: Given a vector V , ∀v ∈ V, 0 · v = 0. The proof is given
as below: (or check the Theorem 3.1.1 in the textbook)
A8 A6 A8
v + 0 · v = 1 · v + 0 · v = (1 + 0) · v = v

So,
A4 A2 A4 A1 A3
0 = −v + v = −v + (v + 0 · v) = (−v + v) + 0 · v = 0 + 0 · v = 0 · v + 0 = 0 · v

2. First, we show that the 0V is in the kernel of L.

L(0V ) = L(0 · 0V ) = L((1 + −1) · 0V ) = L(1 · 0V + −1 · 0V )

= 1 · L(0V ) + (−1) · L(0V ) = (1 + (−1))L(0V ) = 0 · L(0V ) = 0W

Thus, 0V ∈ ker(L)

2
 3. Suppose that there is another vector v̂ ∈ ker(L), but v ̸= 0V . We know that L is one-to-one, i.e.,
∀v1 , v2 ∈ V , L(v1 ) = L(v2 ) ⇒ v1 = v2 . Let v3 = v1 + v̂, by definition of linear transformation, we
have
A3
L(v3 ) = L(v1 + v̂) = L(v1 ) + L(v̂) = L(v1 ) + 0W = L(v1 )

and v3 ̸= v1 since suppose v3 = v1 , we will have

v3 = v1 = v1 + v̂ ⇒ 0 = −v1 + v1 = −v1 + (v1 + v̂) = (−v1 + v1 ) + v̂ = 0 + v̂ = v̂

, this implies 0 = v̂, which is a contradiction.

So, we get a contradiction that L(v3 ) = L(v1 ) but v3 ̸= v1 if there is another vector v̂ ∈ ker(L),
where v̂ ̸= 0
4. In conclusion, we have ker(L) = {0V }.

Proof of if part (⇐)

1. Proof by contradiction. Suppose not, that is, there is a linear transformation L, ker(L) = {0V } but L
is not one-to-one.
2. Since L is not one-to-one, there are 2 vectors v1 , v2 , L(v1 ) = L(v2 ) but v1 ̸= v2 , then we have:

0W = L(0V ) = L(v1 + (−v1 )) = L(v1 ) + L(−v1 ) = L(v2 ) + L(−v1 ) = L(v2 + (−v1 )) (1)

3. Is it possible that v2 + (−v1 ) = 0V ? Suppose v2 + (−v1 ) = 0V , then we have

v2 = v2 + (−v1 ) = 0V ⇒ v2 + (−v1 ) + v1 = 0V + v1 = v1

which is a contradiction. (Or you can use Theorem 3.1.1 in the textbook: If v2 + (−v1 ) = 0V then
v2 = −(−v1 ) = −1 · (−1 · v1 ) = 1 · v1 = v1 , which is a contradiction as well.)
4. By (1), we have

0W = L(v2 + (−v1 )) ̸= 0W

since v2 + (−v1 ) ∈
/ ker(L). This is a contradiction. So, if ker(L) = {0V }, L must be one-to-one.

Question 22
1. To show L maps R3 onto R3 , we need to show L(R3 ) = R3 . To show L(R3 ) = R3 , we need to show 2
things: L(R3 ) ⊇ R3 , L(R3 ) ⊆ R3
 T
2. L(R3 ) ⊆ R3 : ∀x ∈ R3 , let x = x1 x2 x3
 
x1
L(x) =  x1 + x2  ∈ R3
x1 + x2 + x3

So, we have L(R3 ) ⊆ R3 .

 T
3. L(R3 ) ⊇ R3 : ∀y ∈ R3 , let y = y1 y2 y3 , can we find a x ∈ R3 such that L(x) = y? To find such
a x, it is equivalent to solve the linear system:

x1 = y1

x1 + x2 = y2

x1 + x2 + x3 = y3


3
 This linear system can be easily solved:

x 1 = y 1

x2 = y2 − x1 = y2 − y1

x3 = y3 − x1 − x2 = y3 − y1 − (y2 − y1 ) = y3 − y2


Since we can find such a x for any y ∈ R3 , we have L(R3 ) ⊇ R3

4. By the previous discussions, we have L(R3 ) = R3 , which implies L maps R3 onto R3 .

Question 23
1. From question 21, we know that L is one-to-one if and only if ker(L) = {0V }.
2. From question 22, we know that L maps R3 onto R3 if and only if L(R3 ) = R3

For (a) in Question 17

1. By question 17, we know that ker(L) = {0}. By question 21, we can conclude that L is one-to-one.
2. By question 17, we know that L(R3 ) = R3 . Bu question 22, we can conclude that L maps R3 onto R3 .

For (b) in Question 17

1. By question 17, we know that ker(L) ̸= {0}. By question 21, we can conclude that L is not one-to-one.
2. By question 17, we know that L(R3 ) ̸= R3 . Bu question 22, we can conclude that L does not map R3
onto R3 .

For (c) in Question 17

1. By question 17, we know that ker(L) ̸= {0}. By question 21, we can conclude that L is not one-to-one.
2. By question 17, we know that L(R3 ) ̸= R3 . Bu question 22, we can conclude that L does not map R3
onto R3 .

Section 4.2
Question 7
(a)
1. For I(e1 ) = e1 , we write e1 = a1 y1 + a2 y2 + a3 y3 :
       
1 1 1 1
0 = a1 1 + a2 1 + a3 0
0 1 0 0

Solving for a1 , a2 , and a3 , we get

(a1 , a2 , a3 ) = (0, 0, 1).

2. For I(e2 ) = e2 , we write e2 = b1 y1 + b2 y2 + b3 y3 :

       
0 1 1 1
1 = b1 1 + b2 1 + b3 0
0 1 0 0

Solving for b1 , b2 , and b3 , we get

(b1 , b2 , b3 ) = (0, 1, −1).

4
 3. For I(e3 ) = e3 , we write e3 = c1 y1 + c2 y2 + c3 y3 :
       
0 1 1 1
0 = c1 1 + c2 1 + c3 0
1 1 0 0

Solving for c1 , c2 , and c3 , we get

(c1 , c2 , c3 ) = (1, −1, 0).

4. Thus, the coordinates with respect to y1 , y2 , y3 are:

L(e1 ) = (0, 0, 1), L(e2 ) = (0, 1, −1), L(e3 ) = (1, −1, 0).

(b)
Find a matrix A such that Ax is the coordinate vector of x with respect to {y1 , y2 , y3 }.
 
0 0 1
A = 0 1 −1
1 −1 0
note: Theorem 4.2.1

Question 8
(a)
L(y1 ) = (1 + 0 + 0)y1 + (2 + 0)y2 − (0 + 0)y3 = 1y1 + 2y2 + 0y3
L(y2 ) = (0 + 1 + 0)y1 + (0 + 0)y2 − (2 + 0)y3 = y1 + 0y2 − 2y3
L(y3 ) = (0 + 0 + 1)y1 + (0 + 1)y2 − (0 + 1)y3 = 1y1 + 1y2 − 1y3
 
1 1 1
Matrix A = 2 0 1
0 −2 −1

(b)
x = c1 y1 + c2 y2 + c3 y3
• (i)
1.        
7 1 1 1
5 = c1 1 + c2 1 + c3 0
2 1 0 0
⇒ (c1 , c2 , c3 ) = (2, 3, 2)
2.  
  y1
L(y1 ) = 1 2 0 y2 
y3
 
  y1
L(y2 ) = 1 0 −2 y2 
y3
 
  y1
L(y3 ) = 1 1 −1 y2 
y3

5
 3.
x = c1 y1 + c2 y2 + c3 y3

L(x) = c1 L(y1 ) + c2 L(y2 ) + c3 L(y3 )

= 2L(y1 ) + 3L(y2 ) + 2L(y3 )
= 2(y1 + 2y2 ) + 3(y1 − 2y3 ) + 2(y1 + y2 − y3 )
= 7y1 + 6y2 − 8y3

• (ii)

1.        
3 1 1 1
2 = c1 1 + c2 1 + c3 0
1 1 0 0
⇒ (c1 , c2 , c3 ) = (1, 1, 1)
2.
x = c1 y1 + c2 y2 + c3 y3

L(x) = c1 L(y1 ) + c2 L(y2 ) + c3 L(y3 )

= L(y1 ) + L(y2 ) + L(y3 )
= (y1 + 2y2 ) + (y1 − 2y3 ) + (y1 + y2 − y3 )
= 3y1 + 3y2 − 3y3

• (iii)
1.        
1 1 1 1
2 = c1 1 + c2 1 + c3 0
3 1 0 0
⇒ (c1 , c2 , c3 ) = (3, −1, −1)
2.
x = c1 y1 + c2 y2 + c3 y3

L(x) = c1 L(y1 ) + c2 L(y2 ) + c3 L(y3 )

= 3L(y1 ) − L(y2 ) − L(y3 )
= 3(y1 + 2y2 ) − (y1 − 2y3 ) − (y1 + y2 − y3 )
= y1 + 5y2 + 3y3

Section 4.3
Question 4
1. The transition matrix V corresponding to a change of basis from {v1 , v2 , v3 } to {e1 , e2 , e3 } is given by,
 
  1 1 0
V = v1 v2 v3 = 1 2 −2
1 0 1

6
 2. Matrix B that represents L with respect to {v1 , v2 , v3 } is B = V −1 AV . we have

1
V −1 = adj(V )
det(V )
 T
2 −3 −2
1
= −1 1 1
1
−2 2 1
 
2 −1 −2
= −3 1 2
−2 1 1

3. Now, we calculate Matrix B

B = V −1 AV
   
−2 1 2 3 −1 −2 1 1 0
=  3 −1 −2 2 0 −2 1 2 −2
2 −1 −1 2 −1 −1 1 0 1
  
0 0 0 1 1 0
= 3 −1 −2 1 2 −2
2 −1 −1 1 0 1
 
0 0 0
= 0 1 0
0 0 1

Question 8
(a)
1. Let A = SΛS −1 , where Λ is a diagonal matrix with diagonal elements λ1 , λ1 , . . . , λn .
Let s1 , s2 , . . . , sn be the column vectors of S.
2. Since A = SΛS −1 , we have,
AS = (SΛS −1 )S = SΛS −1 S = SΛ.
3. The columns of AS are As1 , As2 , . . . , Asn , and the columns of SΛ are λ1 s1 , λ2 s2 , . . . , λn sn , therefore
we have
Asi = λi si for 1 ≤ i ≤ n.

(b)
1. We know that Ak si = λki si for 1 ≤ i ≤ n for all positive integer k. We have the affirmation is true for
k = 1. Then
Ak si = λki si for 1 ≤ i ≤ n for some positive integer k.
2. We have
Ak+1 si = Ak Asi = Ak (λi si ) = λi Ak si = λi λki si = λk+1
i si for all 1 ≤ i ≤ n.

3. Hence the statement is true for all positive integer k. Then

x = α1 s1 + α2 s2 + · · · + αn sn

4. Using the result above we have

Ak x = Ak (α1 s1 +α2 s2 +· · ·+αn sn ) = α1 Ak s1 +α2 Ak s2 +· · ·+αn Ak sn = α1 λk1 s1 +α2 λk2 s2 +· · ·+αn λkn sn
for all positive integer k.

7
(c)
1. Suppose that |λi | < 1 for 1 ≤ i ≤ n, then we have
lim λki = 0 for all 1 ≤ i ≤ n
k→∞

which gives us that given scalars α1 , α2 , . . . , αn ,

2. Then we get
lim αi λki si = 0
k→∞
therefore
lim (α1 λk1 s1 + α2 λk2 s2 + · · · + αn λkn sn ) = 0
k→∞

3. We have that
If x = α1 s1 + α2 s2 + · · · + αn sn , then
Ak x = α1 λk1 s1 + α2 λk2 s2 + · · · + αn λkn sn ,
for all positive integer k.
4. Hence
lim Ak x = lim (α1 λk1 s1 + α2 λk2 s2 + · · · + αn λkn sn ) = 0.
k→∞ k→∞

Question 14
(a)
1. Let A and B be similar matrices. Then there exists an invertible matrix P such that B = P −1 AP .
2. Given a scalar λ, we have
B − λI = P −1 AP − λI
= P −1 AP − λP −1 P
= P −1 AP − λP −1 IP
= P −1 AP − P −1 (λI)P
= P −1 (A − λI)P

3. By definition, this means that B − λI and A − λI are similar. (Note: CH4 p.64)

(b)
1. From the given scalar λ we have
B−λI = P −1 AP −λI = P −1 AP −λP −1 P = P −1 AP −λP −1 IP = P −1 AP −P −1 (λI)P = P −1 (A−λI)P

Thus, A − λI and B − λI are similar,

2. then, we say
det(A − λI) = det(S −1 (B − λI)S)
= det(S −1 ) det(B − λI) det(S)
= det(S −1 ) det(S) det(B − λI)
= det(S −1 S) det(B − λI)
= det(I) det(B − λI)
= det(B − λI).

3. Therefore, det(A − λI) = det(B − λI). Hence Proved.