You are on page 1of 6

MATH136: Linear Algebra 1 Winter 2023

Written Assignment 4 Solutions

Q1. Let A ∈ Mn×n (F).

(a) Assume that λ ∈ F is an eigenvalue for the matrix A with the corresponding eigenvector

x ∈ Fn . Then, by definition, we have that A #»
x = λ #»
x . Then:

(In − A) x
= In #»
x − A #»
x
= #»
x − λ #»
x (by the properties of matrix-vector multiplication)
= #»
x (1 − λ)
= (1 − λ) #»
x (by the properties of scalar multiplication)
Which by definition, means that (1 − λ) is an eigenvalue for the matrix (In − A) with
the corresponding eigenvector of #»
x.

(b) Assume that µ ∈ F is an eigenvalue for the matrix matrix In − A with the corresponding
eigenvector #»
x ∈ Fn . Then, by definition, we have that (In − A) #»
x = µ #»
x . Then:
#» #»
(In − A) x = µ x
=⇒ In #»
x − A #»
x = µ #»
x (by the properties of matrix-vector multiplication)
=⇒ A #»
x = #»
x − µ #»
x
∴ A #»
x = (1 − µ) #»
x (by the properties of scalar multiplication)
For (1 − µ) ∈ F. Thus, by definition, (1 − µ) is an eigenvalue for the matrix A. Say
(1 − µ) = λ ∈ F; thus, there exist λ being the eigenvalue for A and (1 − µ) = λ =⇒
µ = 1 − λ.

(c) Assume that λ ∈ F is an eigenvalue of A, such that ∀λ, |λ| < 1.


=⇒ −1 < λ < 1
=⇒ 0 < 1 − λ < 2.
By part (a), we know that 1 − λ is an eigenvalue for the matrix (In − A). As found
above, the value of 1 − λ is greater than zero and cannot be 0. Therefore, (In − A)
does not have zero as its eigenvalues and is inverible by the invertibility and eigenvalues
theorem.

1
MATH136: Linear Algebra 1 Winter 2023
Written Assignment 4 Solutions

Q2. Let the recursive sequence (a0 , a1 , . . .) be defined by

a0 = 7, a1 = 26 and an = 7an−1 − 10an−2 for n ≥ 2


   
7 −10 #» 26
and let A = and v = .
1 0 7

(a) We prove this using the principals of strong induction.


Base Cases:  
0 #» #»
For n = 0 we have: A v = I2 v =
26
= #»
c a1 a0 which is true.
7
      
#» 7 −10 26 26 · 7 − 70 a
1
For n = 1 we have: A v = = = 2 which is true.
1 0 7 26 a1
Inductive hypothesis:
 
Assume that for all i ∈ Z and k ∈ N with 0 ≤ i ≤ k, we have that A v = i #» ai+1
. Let
ai
   
v = k+1 be true by the IH. We need to show that Ak+1 #»
Ak #»
a a
v = k+2 is also true.
ak ak+1
We get:
Ak+1 #»
v
= AAk #» v
 
a
= A k+1 (by the IH)
ak
  
7 −10 ak+1
=
1 0 ak
 
7ak+1 − 10ak
=
ak+1
 
ak+2
=
ak+1
as desired.
Therefore, by the principals of strong induction, we get that for all integers
 
n #» an+1
n ≥ 0, A v = .
an

(b) Firstly, let us find the eigenvalues of A. We have that,


CA (λ) = det(A − λI2 )
 
7 − λ −10
= det
1 −λ

2
MATH136: Linear Algebra 1 Winter 2023
Written Assignment 4 Solutions

= −λ(7 − λ) + 10
= (λ − 2)(λ − 5).
Therefore, the eigenvalues of A are λ1 = 2 and λ2 = 5.
Now, let us find the eigenspace corresponding to each eigenvalue.
λ1 = 2:    
5 −10 1 −2
A − 2I2 = ∼ .
1 −2 0 0
   
1 −2 2
Therefore, E2 = Null( ) = Span .
0 0 1
λ2 = 5:    
2 −10 1 −5
A − 5I2 = ∼ .
1 −5 0 0
   
1 −5 5
Therefore, E5 = Null( ) = Span .
0 0 1
     

Take v 1 =
2 #»
∈ E2 and v 2 =
5 #» #»
∈ E5 and construct matrices P = [ v 1 | v 2 ] =
2 5
1 1 1 1
and D = diag(λ1 , λ2 ) = diag(2, 5). As det(P ) = −3 ̸= 0 we have that P is invertible
and also, see that
     
2 5 2 0 −1/3 5/3 7 −10
A = P DP −1 = =
1 1 0 5 1/3 −2/1 1 0
as required.
(c) At first, let us find the inverse of the matrix P introduced in part (b). To do so, we
solve
 the following
  system: 
2 5 1 0 1 0 −1/3 5/3

1 1 0 1 0 1 1/3 −2/3
 
−1/3 5/3
where is the inverse of P .
1/3 −2/3
Knowing that A = P DP −1 by part (b), we get that An = P Dn P −1  forall n ∈ N by
the properties of diagonalization. Therefore, An #» v = P Dn P −1 #»
a
v = n+1 by part (a).
an
Thus, we get:
  n   
n #» n −1 #» 2 5 2 0 −1/3 5/3 26
A v = PD P v =
1 1 0 5 1/3 −2/3 7
  n  
2 5 2 0 3
=
1 1 0 5n 4

2 5 3 · 2n
  
= (by the properties of matrix-vector multiplication)
1 1 4 · 5n

3 · 2n+1 + 4 · 5n+1
   
an+1
= = .
3 · 2n + 4 · 5n an

3
MATH136: Linear Algebra 1 Winter 2023
Written Assignment 4 Solutions

Therefore, we conclude that as the two vectors are equal along with their components,
then an = 3 · 2n + 4 · 5n .

4
MATH136: Linear Algebra 1 Winter 2023
Written Assignment 4 Solutions

Q3. Let U and W be subspaces of Fn and let T : Fn → Fn be a linear transformation.

(a) Let us take a non-zero w #» ∈ W with w #» ̸= #»


0 and let S w#» = { #»
u ∈ U : T ( #» #» As
u ) = w}.

U and W are subspaces, by definition, we know that 0 ∈ U . Also, by the definition
#» #»
of linear transformation. T ( 0 ) = 0 . That is, the zero vector will only be mapped to
the zero vector by the function. As w #» is non-zero, and we want T ( #» #» we get that
u ) = w,
#» #»
u ̸= 0 .

Therefore, 0 ∈ / S w#» and thus, S w#» is not a subspace for Fn by definition.

(b) Let SW = { #»u ∈ U : T ( #»


u ) ∈ W }.
#» #»
Firstly, notice that 0 ∈ U . Take, #» u = 0 ∈ U . Then, by the definition of the linear
#» #»
transformation, T ( #»
u) = T(0) = 0 ∈ W.


∴ 0 ∈ SW .

Secondly, take #» x , #»
y ∈ SW . That is, #»
x , #»
y ∈ U and T ( #»x ), T ( #»
y ) ∈ W . Let c ∈ F.
#» #»
By the properties of subspaces, we get that (c x + y ) ∈ U . Similarily, we have that
cT ( #»
x ) + T ( #»
y ) ∈ W . As the transformation is linear, we can say that T (c #»
x + #»
y ) ∈ W.
Combining the two result, we get that

(c #»
x + #»
y ) ∈ SW .

Thus, by definition, we conclude that SW = { #»


u ∈ U : T ( #»
u ) ∈ W } is a subspace for Fn .

5
MATH136: Linear Algebra 1 Winter 2023
Written Assignment 4 Solutions

#» #» #»
Q4. Let A ∈ Mm×n (F), { b 1 , . . . , b k } ⊆ Fm and #» x i be a solution to A #»
x = b i for i = 1, . . . , k.
#» #»
(a) Assume that { b 1 , . . . , b k } is linearly independent. Equivalently, for ci ∈ F with 1 ≤ i ∈
#» #» #»
N ≤ k ∈ N, we have that c1 b 1 + · · · + ck b k = 0 =⇒ c1 = · · · = ck = 0. Knowing that

A #»
x i = b i , we have
#» #» #»
c1 b 1 + · · · + ck b k = c1 (A #»
x 1 ) + · · · + ck (A #»
x i ) = 0 =⇒ c1 = · · · = ck = 0
=⇒ c1 A = · · · = ck A = B
where B ∈ Mm×n (F) is the zero matrix. It is clear that ∀ #» y ∈ Fn , B #»
y = 0 #»
y and thus,
λ = 0 is an eigenvalue for the zero matrix B. Thus, we get:
c1 (A #»
x 1 ) + · · · + ck (A #»
x i ) = B #»
x 1 + · · · + B #» x k = λ #»
x 1 + · · · + λ #»
x k where λ = 0. Thus, by
#» #»
the linearity of { b 1 , . . . , b k } we can coclude that, for di ∈ F with 1 ≤ i ∈ N ≤ k ∈ N,

d1 #»
x 1 + · · · + dk #»
x k = 0 =⇒ d1 = · · · = dk = λ = 0.
Therefore, by definition, the set { #» x 1 , . . . , #»
x k } is also linearly independent.

(b) Assume that rank(A) = n and { #» x 1 , . . . , #»


x k } is linearly independent. We prove the
#» #»
statement by contradiction and thus assume that { b 1 , . . . , b k } is linearly dependent.
Equivalently, for ci ∈ F with 1 ≤ i ∈ N ≤ k ∈ N, we have:
#» #» #»
c1 b 1 + · · · + ck b k = 0 =⇒ c1 , . . . , ck are not all zero.
Without the loss of generality, take c1 ̸= 0.
#» #» #»
∴ −c1 b 1 = c2 b 2 + · · · + ck b k
#» #» −ck #»
=⇒ b 1 = −c c1 b 2 + · · · + c1 b k
2


=⇒ A #»
x1 = −c2 #»
c1 A x 2 + ··· + −ck #»
c1 A x k (A #»
x i = b i)

=⇒ A(− #» x 1 + −c 2 #» −ck #»
c1 x 2 +· · ·+ c1 x k ) = 0 (by the properties of matrix-vector multiplication).
As rank(A) = n,by the system-rank theorem, the above equation has the unique solution,

(− #»
x 1 + −c 2 #» −ck #»
c1 x 2 + · · · + c1 x k ) = 0 .
Note that as the coefficient of #» x 1 is −1 ̸= 0, we can already see that the set { #»
x 1 , . . . , #»
x k}
is NOT linearly independent by definition. This contradicts our assumption of { #» x 1 , . . . , #»
x k}
#» #»
being linearly independant and therefore, by contradiction, the set { b 1 , . . . , b k } must
be linearly independent.

You might also like