MAT3701/201/1/2016

Tutorial Letter 201/1/2016

LINEAR ALGEBRA

MAT3701

Semester 1

Department of Mathematical Sciences

This tutorial letter contains

solutions to Assignment 01.

BAR CODE

university
Learn without limits.
Open Rubric of south africa
 ASSIGNMENT 01
Solution
UNIQUE ASSIGNMENT NUMBER: 804531

Please note that we will only mark a selection of the questions. It is therefore in your own interest to
do all the questions. The fact that a question is not marked does not mean that it is less important
than one that is marked. We try to cover the whole syllabus over the two semesters (4 assignments)
and to use these assignments to help you prepare for the exam. It is therefore good practice to work
through a complete set of four assignments for a given year, and for this reason the assignments and
worked solutions of previous years are made available on myUnisa under Additional Resources – see
also the letter MAT3701 Exam Preparation under Additional Resources.

Worked solutions to all the questions will be sent to all students and made available on myUnisa
shortly after the due date. Your answers to the assignment questions must be fully motivated.

For this assignment, questions 3, 4, 6, and 7 will be marked.

Question 1

This question was not marked.

It is still important to work through the solutions and compare them with your own attempts.

Consider the vector space V = C 2 with scalar multiplication over the real numbers R. Let

U = {(z1 , z2 ) : z1 = iz2 and z 2 = −z2 }.

(a) Show that U is a subspace of V.

(b) Find a basis for U.

(c) Determine whether V = U ⊕ W, where

W = {(z1 , z2 ) : z1 = iz2 and z 2 = z2 }.

Solution

(a) One approach is to use the Subspace Test (SG, p. 10). Since this is not required, we may instead
show that U has a spanning set and therefore is a subspace. This can be done as follows:

z 2 = −z2 ⇒ z2 imaginary ⇒ z2 = ai, where a ∈ R

thus
U = {(−a, ai) : a ∈ R} = {a(−1, i) : a ∈ R} = span{(−1, i)},
and therefore it follows that U is a subspace of V .

2
 MAT3701/201

(b) It follows from (a) that α = {(−1, i)} is a basis for U .

(c) Similarly as in (a), it follows that

z 2 = z2 ⇒ z2 real ⇒ z2 = a, where a ∈ R

thus
U = {(ia, a) : a ∈ R} = {a(i, 1) : a ∈ R} = span{(i, 1)}.
It follows that U + W = span{(−1, i), (i, 1)}, which is 2-dimensional over R, and since V is
4-dimensional (SG: Section 2.5.2, Example 2(b)), we have V 6= U ⊕ W .

Question 2

This question was not marked.

It is still important to work through the solutions and compare them with your own attempts.

Let V = U ⊕ W, where V is a finite-dimensional vector space over F and U and W are subspaces of
V. If U1 and U2 are subspaces of U and W1 and W2 are subspaces of W, show that
(a) U1 ⊕ W1 is a subspace of V ;
(b) (U1 ⊕ W1 ) ∩ (U2 ⊕ W2 ) = (U1 ∩ U2 ) ⊕ (W1 ∩ W2 ).

Solution
(a) Use the Subspace Test: Let u01 , u001 ∈ U1 and w10 , w100 ∈ W1 and a ∈ F . Then

ST1: 0 = 0 + 0 ∈ U1 ⊕ W1 since 0 ∈ U1 and 0 ∈ W1 .

ST2:

u01 +w10 ∈ U1 ⊕W1 and u001 +w100 ∈ U1 ⊕W1 ⇒ (u01 +w10 )+(u001 +w100 ) = (u01 +u001 )+(w10 +w100 ) ∈ U1 ⊕W1

since (u01 + u001 ) ∈ U1 and (w10 ) + w100 ) ∈ W1 . Thus ST2 holds.

ST3
u01 + w10 ∈ U1 ⊕ W1 and a ∈ F ⇒ a(u01 + w10 ) = au01 + aw10 ∈ U1 ⊕ W1
since au01 ∈ U1 and aw10 ∈ W1 . Thus ST3 holds, and it follows that U1 ⊕ W1 is a subspace of V .

(b)
(U1 ∩ U2 ) ⊕ (W1 ∩ W2 ) ⊆ U1 ⊕ W1
since U1 ∩ U2 ⊆ U1 and W1 ∩ W2 ⊆ W1 . Similarly

(U1 ∩ U2 ) ⊕ (W1 ∩ W2 ) ⊆ U2 ⊕ W2 .

Thus
(U1 ∩ U2 ) ⊕ (W1 ∩ W2 ) ⊆ (U1 ⊕ W1 ) ∩ (U2 ⊕ W2 ).

3
 Conversely

v ∈ (U1 ⊕ W1 ) ∩ (U2 ⊕ W2 ) ⇒ v = u1 + w1 = u2 + w2 , where ui ∈ Ui , wi ∈ Wi , 1 ≤ i ≤ 2

⇒ u1 = u2 and w1 = w2 , since V = U ⊕ W and u1 , u2 ∈ U, w1 , w2 ∈ W
⇒ v ∈ (U1 ∩ U2 ) ⊕ (W1 ∩ W2 ).

Thus (U1 ⊕ W1 ) ∩ (U2 ⊕ W2 ) ⊆ (U1 ∩ U2 ) ⊕ (W1 ∩ W2 ), and the result follows.

Question 3

Let U and W be subspaces of a finite-dimensional vector space V over F such that

dim(U ) = dim(V ) − 1. Prove that

dim(U ∩ W ) = dim(W ) or dim(U ∩ W ) = dim(W ) − 1.

[15]
Solution

Suppose dim(U ∩W ) 6= dim(W ). Then W * U , and therefore V = U +W since dim (U ) = dim(V )−1.
It follows that

dim(V ) = dim (U + W ) = dim (U ) + dim (W ) − dim (U ∩ W )

= dim(V ) − 1 + dim (W ) − dim (U ∩ W ) ,

so
dim(U ∩ W ) = dim(W ) − 1.

Question 4

Consider the vector space V = C 2 with scalar multiplication over the real numbers R. Let T : V → V
be the linear operator defined by     
z1 1 i z1
T = .
z2 −i 1 z2

(a) Find a basis for R(T ). (10)

(b) Find a basis for N (T ). (10)
(c) Determine whether V = R(T ) ⊕ N (T ). (8)
(d) Verify that dim(V )=rank(T ) + nullity(T ). (2)
[30]
Solution
 
1 i
Let A = .
−i 1

4
 MAT3701/201

(a) Since         
1 i 0 0
β = v1 = , v2 = , v3 = , v4 =
0 0 1 i

is a basis for V (SG: Example 2, p.14),

R (T ) = span {T (v1 ) , T (v2 ) , T (v3 ) , T (v4 )}
= span {Av1 , Av2 , Av3 , Av4 }
       
1 i i −1
= span , , ,
−i 1 1 i
   
1 i
= span , .
−i 1

Thus    
1 i
α= ,
−i 1

is a basis for R (T ) since it is linearly independent over R.

 
z1
(b) Let v = , z1 , z2 complex numbers, then
z2
v ∈ N (T ) ⇔ T (v) = 0
⇔ Av
 = 0    
1 i z1 0
⇔ =
−i 1 z2 0
   
z1 + iz2 0
⇔ =
−iz1 + z2 0

z1 + iz2 = 0
⇔
−iz1 + z2 = 0
⇔ z1 + iz2 = 0
⇔ z2 = iz1 .

Let z1 = a + ib, where a, b ∈ R. Then

         
z1 1 a + ib 1 i
v= = z1 = =a +b .
iz1 i ia − b i −1

A basis for N (T ) is therefore    

1 i
β= ,
i −1
since it is linearly independent over R.
(c) Since α ∪ β is a basis for V (show), it follows that V = R(T ) ⊕ N (T ).
(d) Since dim(V ) = 4 and rank(T )= nullity(T ) = 2 from (a) and (b), it follows that
dim(V )=rank(T ) + nullity(T ).

5
Question 5

This question was not marked.

It is still important to work through the solutions and compare them with your own attempts.
 
1 i
Let A = , and let T : M2×2 (C) → M2×2 (C) be the linear operator defined by T (X) = AX.
−i 1
Determine whether or not T is diagonalizable. If it is, find a basis β for M2×2 (C) such that [T ]β is a
diagonal matrix.

Solution
Let         
1 0 0 1 0 0 0 0
β= A11 = , A12 = , A21 = , A22 = .
0 0 0 0 1 0 0 1
Then
 
1 0
T (A11 ) = AA11 = = 1 · A11 + 0 · A12 − i · A21 + 0 · A22
−i 0
 
0 1
T (A12 ) = AA12 = = 0 · A11 + 1 · A12 + 0 · A21 − i · A22
0 −i
 
i 0
T (A21 ) = AA21 = = i · A11 + 0 · A12 + 1 · A21 + 0 · A22
1 0
 
0 i
T (A22 ) = AA22 = = 0 · A11 + i · A12 + 0 · A21 + 1 · A22
0 1

so that  
1 0 i 0
 0 1 0 i 
[T ]β = 
 −i

0 1 0 
0 −i 0 1

6
 MAT3701/201

and
1−λ 0 i 0
  0 1−λ 0 i
det [T ]β − λI =
−i 0 1−λ 0
0 −i 0 1−λ
1−λ 0 i 0 1−λ i
= (1 − λ) 0 1−λ 0 + i −i 0 0
−i 0 1−λ 0 −i 1 − λ
(expanding along the first row)
1−λ i 1−λ i
= (1 − λ)2 −
−i 1 − λ −i 1 − λ
(expanding along the second row of both)
 1−λ i
(1 − λ)2 − 1

=
−i 1 − λ
2
(1 − λ)2 − 1

=

2
= λ2 − 2λ
= λ2 (λ − 2)2 .

Therefore the eigenvalues of T are 0 and 2, both with multiplicity 2.

   
1 0 i 0 1 0 i 0
 0 1 0 i   →  0 1 0 i ,
 
[T ]β − 0I = 
 −i (...i)
0 1 0   0 0 0 0 
0 −i 0 1 0 0 0 0
so  
n − rank [T ]β − 0I = 4 − 2 = 2 = algebraic multiplicity of 0,
and    
−1 0 i 0 1 0 −i 0
 0 −1 0 i 
  0 1 0 −i 
[T ]β − 2I = 
 −i →  (...ii)
0 −1 0   0 0 0 0 
0 −i 0 −1 0 0 0 0
so that  
n − rank [T ]β − 2I = 4 − 2 = 2 = algebraic multiplicity of 2.
Since the geometric and algebraic multiplicities are equal for both eigenvalues, it follows that T is
diagonalizable.
From (i),    

 i 0  
0  i 
    
E0 [T ]β = span  , 
 −1  


 0 
0 −1
 

so that    
i 0 0 i
E0 (T ) = span ,
−1 0 0 −1

7
with basis    
i 0 0 i
β1 = ,
−1 0 0 −1
and, from (ii),    

 i 0 

0 ,  i
     
E2 [T ]β = span  


 1   0 

0 1
 

so that
   
i 0 0 i
E2 (T ) = span ,
1 0 0 1
with basis    
i 0 0 i
β2 = , .
1 0 0 1
Therefore        
i 0 0 i i 0 0 i
β = β1 ∪ β2 = , , ,
−1 0 0 −1 1 0 0 1
is a diagonalizing basis such that  
0 0 0 0
 0 0 0 0 
[T ]β = 
 0
.
0 2 0 
0 0 0 2

Question 6
  
a b
Let T : M2×2 (R) → M2×2 (R) denote the projection on W = : b − c = 0 along U = span
  c d
0 1
.
−1 0

       
1 0 0 1 0 0 0 0
(a) Find the matrix representation of T with respect to β = , , , .
0 0 0 0 1 0 0 1
(18)
 
a b
(b) Find a formula for T expressed as a matrix in terms of a, b, c, d. (7)
c d

[25]

8
 MAT3701/201

Solution
(a) A typical element in W is of the form  
a b
b d
and a typical element in U is of the form
 
0 k
−k 0

where k is a scalar. Therefore a typical element in W + U is of the form

 
a b+k
.
b−k d

Now,    
a b+k 0 1
=
b−k d 0 0
yields a = d = 0, b + k = 1 and b − k = 0. So k = b = 21 from which we obtain that
0 12 1
     
0 1 0 2
= 1 +
0 0 2
0 − 12 0

with
1 1
   
0 0
1
2 ∈W and 2 ∈ U.
2
0 − 12 0
Similarly

1
0 − 21
     
0 0 0 2
= 1 + 1 .
1 0 2
0 2
0
The other two elements of β lie in W , so since T is the projection on W along U , we have
   
1 0 1 0
T =
0 0 0 0

0 12
   
0 1
T = 1
0 0 2
0

0 21
   
0 0
T = 1
1 0 2
0
   
0 0 0 0
T = .
0 1 0 1

Since β is the standard basis for M2×2 (R) , the matrix of T with respect to β is given by
 
1 0 0 0
 0 1 1 0 
2 2
[T ]β = 
 0 1 1 0 .

2 2
0 0 0 1

9
 (b) Since          
a b 1 0 0 1 0 0 0 0
=a +b +c +d ,
c d 0 0 0 0 1 0 0 1
we have
    
a b 1 0 0 1
T = aT + bT
c d 0 0 0 0
   
0 0 0 0
+cT + dT
1 0 0 1

0 21 0 21
       
1 0 0 0
= a +b 1 +c 1 +d
0 0 2
0 2
0 0 1

a b+c
 
= 2 .
b+c
2
d

Question 7

Let T : M4×4 (C) → M4×4 (C) be the linear operator over C defined by T (X) = AX, where
 
0 0 0 0
 1 0 1 0 
A=  0 1
,
0 0 
0 0 0 1
and let W be the T –cyclic subspace of M4×4 (C) generated by A.
(a) Find the T –cyclic basis for W. (10)
(b) Find the characteristic polynomial of TW . (3)
(c) Determine whether TW is one-to-one. (4)
(d) Determine whether TW is onto. (3)
(e) For each eigenvalue of TW , find a corresponding eigenvector expressed as a linear combination
of the T -cyclic basis for W. (10)
[30]
Solution
(a)
   
0 0 0 0 0 0 0 0
 1 0
0 
 1  0 1 0 0 
A = 
 0 , T (A) = A2 =  ,
1
0  0  1 0 1 0 
0 1 0 0 0 0 0 1
 
0 0 0 0
 1 0 1 0 
T 2 (A) = T (T (A)) = T A2 = A3 = 


 = A.
 0 1 0 0 
0 0 0 1

10
 MAT3701/201

Since A and T (A) are linearly independent and T 2 (A) = A, it follows that the T –cyclic basis
for W is {A, T (A)} = {A, A2 } .

(b) Since T 2 (A) = A from (a), we have cTW = (−1)2 (t2 − 1) = t2 − 1.

(c) Since 0 is not a root of cTW , 0 is not an eigenvalue of TW and therefore TW is one-to-one.

(d) Since TW is one-to-one and W is finite-dimensional it follows that TW is onto.

(e) Since from (b),

T 2 (A) − A = (T 2 − I)(A) = (T − I)(T + I)(A) = 0
it follows that (T + I)(A) = T (A) + A is an eigenvector of TW corresponding to λ = 1 and
(T − I)(A) = T (A) − A is an eigenvector of TW corresponding to λ = −1.

Question 8

This question was not marked.

It is still important to work through the solutions and compare them with your own attempts.

 √ √ 
1√ 2 + i 2 −1
(a) Let B = 1 + i 3
 2 1 .
3 − 4i 0 −1

(i) Describe the Gerschgorin discs in which the eigenvalues of B lie.

(ii) Calculate the row and column sums of B.
 
0.4 0.7
(b) Let A = .
0.6 0.3

(i) Find the smallest disc centered at the origin of the complex plane in which the eigenvalues
of A lie.
(ii) Find an eigenvalue of A on the circumference of your disc in (i), and a corresponding
eigenvector.
(iii) Find lim Am .
m→∞

Solution

(a)(i)
√ √
C1 = {z ∈ C : |z − 1| ≤ | 2 + i 2| + | − 1| = 3}
√
C2 = {z ∈ C : |z − 2| ≤ |1 + i 3| + |1| = 3}
C2 = {z ∈ C : |z + 1| ≤ |3 − 4i| + |0| = 5}.

11
(ii)

ρ(B) = max{ρ1 (B), ρ2 (B), ρ3 (B)}

√ √ √
= max{|1| + | 2 + i 2| + | − 1|, |1 + i 3| + |2| + |1|, |3 − 4i| + |0| + | − 1|}
= max{4, 5, 6}
=6
ν(B) = max{ν1 (B), ν2 (B), ν3 (B)}
√ √ √
= max{|1| + |1 + i 3| + |3 − 4i|, | 2 + i 2| + |2| + |0|, | − 1| + |1| + | − 1|}
= max{8, 4, 3}
= 8.

(b)(i) It is the disc with radius min{ρ(A), ν(A)} =min{1.1, 1} = 1.

(ii) Since A is a transition matrix, λ = 1 is always an eigenvalue, and it lies on the circumference
of the disc in b(i). To find a corresponding eigenvector, we solve the following homogeneous linear
system with coefficient matrix:
   
−0.6 0.7 1 7
(A − I) = of which v =
0.6 −0.7 13 6

is a solution, and therefore an eigenvector corresponding to λ = 1.

(iii) Since A is a regular transition matrix and v in b(ii) is an eigenvector of A corresponding to λ = 1
which is also a probability vector, it follows that
 
m 1 7 7
lim A = .
m→∞ 13 6 6

Total [100]

12
 MAT3701/202/1/2016

Tutorial Letter 202/1/2016

LINEAR ALGEBRA

MAT3701

Semester 1

Department of Mathematical Sciences

This tutorial letter contains

solutions to Assignment 02.

BAR CODE

university
Learn without limits.
Open Rubric of south africa
 ASSIGNMENT 02
Solution
UNIQUE ASSIGNMENT NUMBER: 690237

Please note that we will only mark a selection of the questions. It is therefore in your own best
interest to attempt all the questions. The fact that a question is not marked does not mean that it
is less important than one that is marked. We try to cover the whole syllabus over the two semesters
(4 assignments) and to use the assignments to help you prepare for the exam. It is therefore good
practice to work through a complete set of four assignments for a given year, and for this reason the
assignments and worked solutions of previous years are made available on myUnisa under Additional
Resources – see also the letter MAT3701 Exam Preparation under Additional Resources.

Worked solutions for this assignment will be made available on myUnisa shortly after the due date.
Your answers to the assignment questions should be fully motivated.

For this assignment, Questions 2, 5, 6, and 8 will be marked.

Question 1

This question was not marked.

It is still important to work through the solutions and compare them with your own attempts.

Consider the inner product space P2 (R) over R with h·, ·i defined by

hg, hi = g (a) h (a) + g(b)h (b) + g (c) h (c)

and let β = {fa , fb , fc } be the set of Lagrange polynomials associated with the distinct real numbers
a, b, and c.

(a) Show that β is linearly independent.

(b) Deduce from (a) that β is a basis for P2 (R) .

(c) Show that β is orthonormal with respect to h·, ·i .

(d) Show that, with respect to h·, ·i , the orthogonal projection P : P2 (R) → P2 (R) on W =
span{fa , fb } is given by P (g) = g (a) fa + g (b) fb .

(e) Find [P ]β .

Solution

2
 MAT3701/202

(a) For r1 , r2 , r3 ∈ R,

r1 fa (a) + r2 fb (a) + r3 fc (a) = 0

r1 fa + r2 fb + r3 fc = 0 ⇒ r1 fa (b) + r2 fb (b) + r3 fc (b) = 0

r1 fa (c) + r2 fb (c) + r3 fc (c) = 0


r1 · 1 + r2 · 0 + r3 · 0 = 0

⇒ r1 · 0 + r2 · 1 + r3 · 0 = 0

r1 · 0 + r2 · 0 + r3 · 1 = 0


⇒ r1 = r2 = r3 = 0
⇒ β is linearly independent.

(b) From (a), β consists of three linearly independent vectors, and since P2 (R) is of dimension 3, it
follows that β is a basis for P2 (R) .

(c)

hfa , fa i = fa (a) fa (a) + fa (b) fa (b) + fa (c) fa (c)

= 1.1 + 0.0 + 0.0
= 1,

and similarly hfb , fb i = hfc , fc i = 1. Also,

hfa , fb i = fa (a) fb (a) + fa (b) fb (b) + fa (c) fb (c)

= 1.0 + 0.1 + 0.0
= 0,

and similarly hfa , fc i = hfb , fc i = 0. Thus {fa , fb , fc } is an orthonormal basis for P2 (R).

(d) By definition,

P (g) = hg, fa i fa + hg, fb i fb

= (g (a) fa (a) + g (b) fa (b) + g (c) fa (c)) fa
+ (g (a) fb (a) + g (b) fb (b) + g (c) fb (c)) fb
= (g (a) · 1 + g (b) · 0 + g (c) · 0) fa + (g (a) · 0 + g (b) · 1 + g (c) · 0) fb
= g (a) fa + g (b) fb .

(e) P (fa ) = fa (a) fa + fa (b) fb = 1 · fa + 0 · fb + 0 · fc

P (fb ) = fb (a) fa + fb (b) fb = 0 · fa + 1 · fb + 0 · fc
P (fc ) = fc (a) fa + fc (b) fb = 0 · fa + 0 · fb + 0 · fc ,
thus  
1 0 0
[P ]β =  0 1 0 .
0 0 0

3
Question 2

Given that the system

x − y +z = 0
x+y = 2
x − y +z = 0
has infinitely many solutions in R3, find the minimal solution. [11]

Solution

Let    
1 −1 1 0
A= 1
 1 0  and b = 2 .

1 −1 1 0
We follow the method outlined in Friedberg: Section 6.3, Example 3.
    
1 −1 1 1 1 1 3 0 3
∗
AA = 1  1 0   −1 1 −1  =  0 2 0 
1 −1 1 1 0 1 3 0 3
so
AA∗ x = b    
3 0 3 x1 0
⇔  0 2 0   x2 = 2 
 
3 0 3 x3 0

3x1 +3x3 = 0 
⇔ 2x2 =2
3x1 +3x3 = 0

 
0
A solution to the system is e.g.  1 , therefore the minimal solution is
0
    
1 1 1 0 1
s = A∗ x =  −1 1 −1   1  =  1 .
1 0 1 0 0

Question 3

This question was not marked.

It is still important to work through the solutions and compare them with your own attempts.

Let M2×2 (C) be the inner product space with inner product defined by hA, Bi = tr (B ∗ A) (see SG:
Example 6, p. 106), and let T : M2×2 (C) → M2×2 (C) be the linear operator defined by T (A) = GA,
where G ∈ M2×2 (C) is a fixed matrix.

4
 MAT3701/202

(a) Show that T ∗ (A) = G∗ A for all A ∈ M2×2 (C) . Thus, T is self-adjoint if G is self-adjoint.
 
1 i
(b) Let G = (hence T is self-adjoint). Find an orthonormal basis for M2×2 (C) consisting
−i 1
of eigenvectors of T.

Solution

(a)

hT ∗ (A) , Bi = hA, T (B)i

= hA, GBi
= tr ((GB)∗ A)
= tr (B ∗ G∗ A)
= hG∗ A, Bi for all A, B ∈ M2×2 (R) .
Therefore
T ∗ (A) = G∗ A for all A ∈ M2×2 (C) .

(b) The standard basis for M2×2 (C) is given by

        
1 0 0 1 0 0 0 0
β = B11 = ; B12 = ; B21 = ; B22 = .
0 0 0 0 1 0 0 1

Now,
    
1 i 1 0 1 0
T (B11 ) = = = B11 + 0 · B12 − i · B21 + 0 · B22
−i 1 0 0 −i 0
    
1 i 0 1 0 1
T (B12 ) = = = 0 · B11 + B12 + 0 · B21 − i · B22
−i 1 0 0 0 −i
    
1 i 0 0 i 0
T (B21 ) = = = i · B11 + 0 · B12 + B21 + 0 · B22
−i 1 1 0 1 0
    
1 i 0 0 0 i
T (B22 ) = = = 0.B11 + i · B12 + 0 · B21 + B22
−i 1 0 1 0 1

So  
1 0 i 0
 0 1 0 i 
[T ]β = 
 −i
.
0 1 0 
0 −i 0 1

5
  
f (λ) = det [T ]β − λ I

1−λ 0 i 0
=
0 1−λ 0 i
−i 0 1−λ 0
0 −i 0 1−λ

1−λ 0 i 0 1−λ i
= (1 − λ)
0 1−λ 0 +i −i 0 0 (expansion along first row)
−i 0 1−λ 0 −i 1−λ

1−λ i 1−λ i
= (1 − λ)2 −
−i 1−λ −i 1−λ
2
= (1 − λ)2 − 1


= λ2 (λ − 2)2 .
 
E0 [T ]β : Solve the homogeneous system with coefficient matrix.
   
1 0 i 0 1 0 i 0
 0 1 0 i 
  0 1 0 i 
[T ]β − 0 · I = 
 −i → 
0 1 0   0 0 0 0 
0 −i 0 1 0 0 0 0

Therefore    

 1 0 

0 ,  1
     
E0 [T ]β = span   (check)


 i   0 

0 i
 

so    
1 0 0 1
E0 (T ) = span , ,
i 0 0 i

which is orthogonal since the Froberius inner product is given by

hA, Bi = tr (B ∗ A) = a11 b11 + a12 b12 + a21 b21 + a22 b22 , where A = (aij ) and B = (bij ) .
 
E2 [T ]β : Solve the system with coefficient matrix
   
−1 0 i 0 −1 0 i 0
[T ]β − 2 · I = 
 0 −1 0  →  0 −1 0 i 
i   
 −i 0 −1 0   0 0 0 0 
0 −i 0 −1 0 0 0 0

6
 MAT3701/202

Therefore    

 i 0 

0   i
     
E2 [T ]β = span  ,  0
 (check)
 1
  

0 1
 

so    
i 0 0 i
E2 (T ) = span , ,
1 0 0 1

which is orthogonal with respect to the given inner product. It follows that
        
1 1 0 1 0 1 1 i 0 1 0 i
γ= √ , √ , √ , √
2 i 0 2 0 i 2 1 0 2 0 1

is an orthonormal basis for M2×2 (C) consisting of eigenvectors of T.

Question 4

This question was not marked.

It is still important to work through the solutions and compare them with your own attempts.

Let V = W ⊕ W ⊥, where V is a finite-dimensional inner product space over F and W is a subspace

of V. Define U : V → V by U (v1 + v2 ) = v1 − v2 , where v1 ∈ W and v2 ∈ W ⊥ . Prove that U is a
self-adjoint unitary operator.

Solution

Let w1 , w2 ∈ W, w1⊥ , w2⊥ ∈ W ⊥ , and a ∈ F. First we show that U is a linear operator.

U ((w1 + w1⊥ ) + (w2 + w2⊥ )) = U ((w1 + w2 ) + (w1⊥ + w2⊥ ))

= (w1 + w2 ) − (w1⊥ + w2⊥ )
= (w1 − w1⊥ ) + (w2 − w2⊥ )
= U (w1 + w1⊥ ) + U (w2 + w2⊥ ),

and

U (a(w1 + w1⊥ )) = U (aw1 + aw1⊥ )

= aw1 − aw1⊥
= a(w1 − w1⊥ )
= aU (w1 + w1⊥ ).

Thus U is a linear operator.

7
To show that U is self-adjoint, note that

hU (w1 + w1⊥ ), w2 + w2⊥ i = hw1 − w1⊥ , w2 + w2⊥ i

= hw1 , w2 i + hw1 , w2⊥ i − hw1⊥ , w2 i − hw1⊥ , w2⊥ i
= hw1 , w2 i − hw1⊥ , w2⊥ i

and

hw1 + w1⊥ , U (w2 + w2⊥ )i = hw1 + w1⊥ , w2 − w2⊥ i

= hw1 , w2 i − hw1 , w2⊥ i + hw1⊥ , w2 i − hw1⊥ , w2⊥ i
= hw1 , w2 i − hw1⊥ , w2⊥ i.

So hU (w1 + w1⊥ ), w2 + w2⊥ i = hw1 + w1⊥ , U (w2 + w2⊥ )i and thus U is self-adjoint.

Finally, to show that U is unitary, note that

U 2 (w1 + w1⊥ ) = U (w1 − w1⊥ ) = w1 + w1⊥ , so U 2 = IV

and since U is also self-adjoint, U ∗ U = U 2 = IV . Thus U is unitary. It therefore follows that U is a

self-adjoint unitary operator.

Question 5

Eliminate the xy–term in

√
5x2 − 2 3xy + 7y 2 = 1 ... ( ∗ )

by a rotation of the axes, that is, find a rotation matrix P ∈ M2×2 (R) such that (∗), expressed in
terms of x0 , y 0 defined by    0 
x x
=P
y y0
contains no cross term. Express x, y in terms of x0 , y 0 , and state the (counterclockwise) angle of
rotation. [17]

Solution
√  
√

5√ − 3 x
5x2 − 2 3xy + 7y 2 = x
 
y .
− 3 7 y
 √ 
5√ − 3
Let A = , then
− 3 7
√
5−√λ − 3
|A − λI| =
− 3 7−λ
= (5 − λ)(7 − λ) − 3
= λ2 − 12λ + 32
= (λ − 4)(λ − 8),

8
 MAT3701/202

hence the eigenvalues of A are λ1 = 4 and λ2 = 8.

 √   √ 
1√ − 3 1 − 3 √
E4 : A − 4I = →
− 3 3 0 0 R2 + 3R1
so   √ 
1 3
E4 = span .
2 1
 √   √ 
−3√ − 3 − 3 −1 R2 √
E8 : A − 8I = →
− 3 −1 0 0 R1 − 3R2
so   
1 √1
E8 = span .
2 − 3
To ensure that det(P ) = 1, choose the spanning vector of E8 as the first column of P :
 √ 
1 1 3
P = √ .
2 − 3 1

Then P is a rotation matrix since it is orthogonal and det(P ) = 1.

   0 
x x
= P
y y0
 √  0 
1 1 3 x
= √
2 − 3 1 y0
 0 √ 0 
1 x√+ 3y
=
2 − 3x0 + y 0
so √
1 0 3 0 
x= x+ y

2
√ 2
3 0 1 0 
y=− x + 2y 
2
The conic section (∗) in terms of x0 , y 0 becomes

8x02 + 4y 02 = 1.

Since
cos 53 π − sin 53 π
 

P = ,
5 5
sin 3
π cos 3
π
5 π
the angle of rotation is π (or − = −60◦ ).
3 3

9
Question 6

Let V = M2×2 (R) denote the inner product space over R with respect to the Frobenius inner product,
that is,    
z1 z2 y1 y2
, = z1 y1 + z2 y2 + z3 y3 + z4 y4 .
z3 z4 y3 y4
Let T : V → V denote the orthogonal projection on
     
1 0 0 0 1 0 1
W = span (S) , where S = , ,√ .
0 0 0 1 2 1 0
 
z1 z2
(a) Find the formula for T , expressed as a single matrix in terms of z1 , z2 , z3 and z4 .(7)
z3 z4
 
z1 z2
(b) Use T to express A = as A = B + C where B ∈ W and C ∈ W ⊥ . (4)
z3 z4

[11]
Solution

(a) It is easy to check that

     
1 0 0 0 1 0 1
S= , ,√
0 0 0 1 2 1 0
is an orthonormal basis for W.
 
z1 z2
Let A = , and denote the elements of S by S1 , S2 , and S3 , respectively. Then
z3 z4

T (A) = hA, S1 i S1 + hA, S2 i S2 + hA, S3 i S3

= z1 S1 + z4 S2 + √1 (z2 + z3 )S3
2

1
 
z1 (z
2 2
+ z3 )
= 1
(z
2 2
+ z3 ) z4

i.e.
1
   
z1 z2 z1 (z
2 2
+ z3 )
T = 1 .
z3 z4 (z + z3 )
2 2
z4

(b) Since T is an orthogonal projection,

A = T (A) + (I − T )(A) where T (A) ∈ W and (I − T )(A) ∈ W ⊥

thus
1 1
   
z1 (z
2 2
+ z3 ) 0 (z
2 2
− z3 )
B= 1 and C = .
(z + z3 )
2 2
z4 − 21 (z2 − z3 ) 0

10
 MAT3701/202

Question 7

This question was not marked.

It is still important to work through the solutions and compare them with your own attempts.

Find the spectral decomposition of

 
1+i i 0
A= i 1+i 0 .
0 0 1 + 2i

(Write each orthogonal projection as a single matrix.)

Solution
1+i−t i 0
det(A − tI3 ) = i 1+i−t 0
0 0 1 + 2i − t
1+i−t i
= (1 + 2i − t)
i 1+i−t
= (1 + 2i − t)[(1 + i − t)2 − i2 ]
= (1 + 2i − t)2 (1 − t).

So the eigenvalues of A are λ1 = 1 and λ2 = 1 + 2i (with multiplicity two).

   
i i 0 1 1 0 −iR1
E1 : A − I3 = i
 i 0  → 0 0 0
  R2 − R1
0 0 2i 0 0 1 − 21 iR3
so   
 1 1 
E1 = span √  −1  .
 2
0

   
−i i 0 1 −1 0 iR1
E1+2i : A − (1 + 2i) I3 =  i −i 0 → 0 0 0
   R2 + R1
0 0 0 0 0 0
so     
 1 1 0 
E1+2i = span √  1  ,  0  .
 2
0 1


Let
√1 √1 0
 
2 2
P = − √12 √1
2
0 ,
0 0 1

11
then the spectral decomposition of A is
   
1 0 0 0 0 0
A = P  0 0 0  P ∗ + (1 + 2i) P  0 1 0  P ∗
0 0 0 0 0 1
 1   1 
√ √ 0  1
 √21  √1 − √1 0  √12 √1

h i
 √2 2
0
=  − 2  2 2
+ (1 + 2i)  2 0 
0 0 1
0 0 1
 1
− 12 0
  1 1 
2 2 2
0
=  − 12 12 0  + (1 + 2i)  21 12 0 .
0 0 0 0 0 1

Question 8
 
1 1
Let A =  √  and suppose k·k denotes the Euclidean norm.
√
2 − 2
(a) Find kAk , kA−1 k and cond(A). (7)
e such that Ax = b, kbk = 1, and kb − Ae
(b) Suppose that we have vectors x and x xk ≤ 0.001. Use (a)
−1
to determine upper bounds for ke x − A bk (the absolute error) and ke x − A−1 bk / kA−1 bk (the
relative error). (4)
[11]
Solution

(a) √    
∗1 √2 √1 1
√ = 3 −1
A A= ,
1 − 2 2 − 2 −1 3
thus
λ−3 1
det(λI − A∗ A) =
1 λ−3
= (λ − 3)2 − 1
= (λ − 4)(λ − 2),
so
√ 1 2 √
kAk = 4 = 2, A−1 = √ , and cond (A) = kAk · A−1 = √ = 2.
2 2
(b)
1
e − A−1 b = A−1 (Ae
x x − b) ≤ A−1 kAe
x − bk ≤ √ (0.001),
2
and
x − A−1 bk
ke ke
x − xk kAe
x − bk √
= ≤ cond (A) ≤ 0.001 2.
kA−1 bk kxk kbk
Total [50]

12
 MAT3701/201/2/2016

Tutorial Letter 201/2/2016

LINEAR ALGEBRA

MAT3701

Semester 2

Department of Mathematical Sciences

This tutorial letter contains

solutions to Assignment 01.

BAR CODE

university
Define tomorrow. of south africa
 ASSIGNMENT 01
Solution
UNIQUE ASSIGNMENT NUMBER: 692335

Please note that we will only mark a selection of the questions. It is therefore in your own best
interest to do all the questions. The fact that a question is not marked does not mean that it is
less important than one that is marked. We try to cover the whole syllabus over the two semesters
(4 assignments) and to use these assignments to help you prepare for the exam. It is therefore good
practice to work through a complete set of four assignments for a given year, and for this reason the
assignments and worked solutions of previous years are made available on myUnisa under Additional
Resources – see also the letter MAT3701 Exam Preparation under Additional Resources.

Worked solutions to all the questions will be sent to all students and made available on myUnisa
shortly after the due date. Your answers to the assignment questions must be fully motivated.

For this assignment, questions 1, 3, 7 and 8 will be marked.

Question 1
   
1 −2 −i 0
Let A = and B = , and let
1 −1 0 i

W1 = {X ∈ M2×2 (C) : AX = XB}

and
W2 = {X ∈ M2×2 (C) : AX = iX}.

(a) Show that W1 is a subspace of M2×2 (C). (8)

(b) Given that W2 is a subspace of M2×2 (C), find a basis for W1 ∩ W2 . (18)

(c) Explain whether or not M2×2 (C) = W1 ⊕ W2 . (2)

[28]

Solution

(a) ST1: W1 6= φ since e.g. 02×2 ∈ W1 .

ST2: Suppose X1 ∈ W1 and X2 ∈ W1 . Then

A (X1 + X2 ) = AX1 + AX2

= X1 B + X2 B since X1 , X2 ∈ W1
= (X1 + X2 ) B
Thus X1 + X2 ∈ W1 .

2
 MAT3701/201

ST3: Suppose z ∈ C and X ∈ W1 . Then

A (zX) = z (AX)
= z (XB) since X ∈ W1
= (zX) B
Thus zX ∈ W1 .

Since all three conditions are satisfied, W1 is a subspace of M2×2 (C) .

(b) Let X = [X1 X2 ] , where X1 and X2 denote the first and second columns of X respectively.
Then
X ∈ W1 ∩ W2 ⇔ AX = XB and AX = iX  
−i 0
⇔ A [X1 X2 ] = [X1 X2 ] and A [X1 X2 ] = i [X1 X2 ]
0 i
⇔ [AX1 AX2 ] = [−iX1 iX2 ] and [AX1 AX2 ] = [iX1 iX2 ]
⇔ AX1 = −iX1 = iX1 and AX2 = iX2
⇔ X1 = 0 and AX2 = iX2 ...(i)
 
x1
Let X2 = then
x2
    
1 −2 x1 x1
AX2 = iX2 ⇔ =i
1 −1 x2 x2
    
1−i −2 x1 0
⇔ =
1 −1 − i x2 0
1+i
⇔ x1 = (1 + i)x2 (multiply the first row by )
2

Thus    
(1 + i)x2 1+i
X2 = = x2 , x2 ∈ C
x2 1

and therefore  
0 1+i
X = [X1 X 2 ] = x2 , x2 ∈ C
0 1
 
0 1+i
So is a basis for W1 ∩ W2 .
0 1

(c) No, since W1 ∩ W2 6= {0} .

3
Question 2

This question was not marked.

It is still important to work through the solutions and compare them with your own attempts.

Let V = U ⊕ W, where V is a finite-dimensional vector space over F and U and W are subspaces of
V, and let f : U → U and g : W → W be linear operators on U and W respectively. Show that
(a) f ⊕ g : V → V defined by

f ⊕ g(u + w) = f (u) + g(w), where u ∈ U and w ∈ W

is a linear operator on V.
(b) Show that f ⊕ g is one-to-one if and only if both f and g are one-to-one.
(c) Show that f ⊕ g is onto if and only if both f and g are onto.

Solution

(a) Let v1 , v2 ∈ V and a ∈ F . There exist vectors u1 , u2 ∈ U and w1 , w2 ∈ W such that v1 = u1 +w1
and v2 = u2 + w2 .

Now

f ⊕ g(v1 + v2 ) = f ⊕ g((u1 + w1 ) + (u2 + w2 ))

= f ⊕ g((u1 + u2 ) + (w1 + w2 ))
= f (u1 + u2 ) + g(w1 + w2 ), definition of f ⊕ g
= f (u1 ) + f (u2 ) + g(w1 ) + g(w2 ), since f and g are linear operators
= (f (u1 ) + g(w1 )) + (f (u2 ) + g(w2 ))
= f ⊕ g(u1 + w1 ) + f ⊕ g(u2 + w2 ), definition of f ⊕ g
= f ⊕ g(v1 ) + f ⊕ g(v2 )

and

f ⊕ g(av1 ) = f ⊕ g(a(u1 + w1 ))
= f ⊕ g(au1 + aw1 ))
= f (au1 ) + g(aw1 ), definition of f ⊕ g
= af (u1 ) + ag(w1 ), since f and g are linear operators
= a(f (u1 ) + g(w1 ))
= a(f ⊕ g)(u1 + w1 ), definition of f ⊕ g
= a(f ⊕ g)(v1 )

Thus f ⊕ g is a linear operator since both conditions are satisfied.

4
 MAT3701/201

(b) Suppose f and g are both one-to-one. For v = u + w with u ∈ U and w ∈ W ,

f ⊕ g(v) = 0 ⇒ f ⊕ g(u + w) = 0
⇒ f (u) + g(w) = 0, definition of f ⊕ g
⇒ f (u) = g(w) = 0, since f (u) ∈ U and g(w) ∈ W and V = U ⊕ W
⇒ u = w = 0, since f and g are one-to-one
⇒v=0
⇒ f ⊕ g is one-to-one

Conversely, suppose f ⊕ g is one-to-one. For u ∈ U and w ∈ W ,

f (u) = g(w) = 0 ⇒ f ⊕ g(u + w) = f (u) + g(w) = 0

⇒ u + w = 0 since f ⊕ g is one-to-one
⇒ u = w = 0, since u ∈ U and w ∈ W and V = U ⊕ W
⇒ f and g are both one-to-one

(c) Suppose f and g are both onto. Let v = u + w with u ∈ U and w ∈ W . Then there exist
vectors u1 ∈ U and w1 ∈ W such that f (u1 ) = u and g(w1 ) = w. Therefore

f ⊕ g(u1 + w1 ) = f (u1 ) + g(w1 ) = u + w = v

and hence f ⊕ g is onto.

Conversely, suppose f ⊕ g is onto. Let u ∈ U and w ∈ W . Then there exist vectors u1 ∈ U and
w1 ∈ W such that

f ⊕ g(u1 + w1 ) = u + w
⇒f (u1 ) + g(w1 ) = u + w, definition of f ⊕ g
⇒f (u1 ) = u and g(w1 ) = w, since f (u1 ), u ∈ U and g(w1 ), w ∈ W and V = U ⊕ W
⇒f and g are both onto

Question 3

Let V be the vector space C 2 with scalar multiplication over the real numbers R, and let T : V → V
be the linear operator defined by

T (z1 , z2 ) = (z1 + z 1 , z2 − z 2 ).

Determine whether or not T is diagonalizable. If it is, find a basis β for V such that [T ]β is a diagonal
matrix. [15]

Solution

5
According to SG: Example 2(b),

γ = {(1, 0), (i, 0), (0, 1), (0, i)}

is a basis for V . Let us calculate [T ]γ :

T (1, 0) = (2, 0) = 2(1, 0) + 0(i, 0) + 0(0, 1) + 0(0, i)

T (i, 0) = (0, 0) = 0(1, 0) + 0(i, 0) + 0(0, 1) + 0(0, i)
T (0, 1) = (0, 0) = 0(1, 0) + 0(i, 0) + 0(0, 1) + 0(0, i)
T (0, i) = (0, 2i) = 0(1, 0) + 0(i, 0) + 0(0, 1) + 2(0, i)

so  
2 0 0 0
0 0 0 0
[T ]γ = 
0

0 0 0
0 0 0 2
Thus T is diagonalizable and β = γ is a suitable choice for β.

Question 4

This question was not marked.

It is still important to work through the solutions and compare them with your own attempts.

Let T : V → V be a linear operator on a vector space V over F.

(a) Show that T 2 = T ⇔ R (T ) ⊆ N (T − I).

(b) Suppose that V = R4 and F = R. Find a formula for T such that T 2 = T ,

   

 1 0 
1   0 
   
R(T ) = span  0 , 
 (*)

 1 

0 0
 

and N (T ) = R(T )⊥ .

Solution

(a)

T 2 = T ⇔ T 2 (v) = T (v) for all v ∈ V

⇔ T (T (v)) − T (v) = 0 for all v ∈ V
⇔ (T − I)(T (v)) = 0 for all v ∈ V
⇔ T (v) ∈ N (T − I) for all v ∈ V
⇔ R (T ) ⊆ N (T − I)

6
 MAT3701/201

(b)
          

 x 1 x 1 1 x 1 0 

 x2  x2  1 x2  0 
N (T ) =   :   ·   =   ·   = 0
         

 x3 x3 0 x3 1 

x4 x4 0 x4 0
 
  

 x1 

x
  
2
=  x3 
 : x1 + x2 = x3 = 0

 

x4
 
  

 x1 

−x
 
1

=    : x1 , x4 ∈ R

 0  

x4
 
   

 1 0 
   
−1  , 0

= span   0  0 (**)

 
0 1
 

Since V = R(T ) ⊕ R(T )⊥ and N (T ) = R(T )⊥ it follows from (*) and (**) that
       

 1 0 1 0  
1 0 −1   0 
    
β=   0 ,  1 ,  0
    ,  
  0 

 
0 0 0 1
 

is a basis for V. According to (a), T (v) = v for all v ∈ R(T ). Thus it follows from (*) and (**)
that
             
1 1 0 0 1 0 0
 1   1   0   0   −1   0   0 
T 0  =  0  , T  1  =  1  and T =  0  = T  0  =  0 
            

0 0 0 0 0 1 0

Now
       
1 1 1 1
0 1 1 1 −1 1 1
0 = 2 T 0 + 2 T  0  = 2 0
T       

0 0 0 0
       
0 1 1 1
1 1 1 1 −1 1 1
0 = 2 T 0 − 2 T  0  = 2 0
T       

0 0 0 0
       
0 0 0 0
0 0 0 0
T
1 = 1 and T 0 = 0
      

0 0 1 0

7
 So
         
a 1 0 0 0
 b 
 = aT  0  + bT  1 
 + cT  0  + dT  0 
     
T
 c   0 

 0   1   0 
d 0 0 0 1
       
1 1 0 0
a 1  b 1
     0   0 
= +  + c  + d 
2 0  2 0   1   0 
0 0 0 0
 a+b 
2
 a+b 
=  2 
 c 
0

Question 5

This question was not marked.

It is still important to work through the solutions and compare them with your own attempts.

Suppose that A ∈ Mn×n (F ) has two distinct eigenvalues λ1 and λ2 and that dim(Eλ1 ) = n − 1. Prove
that A is diagonalizable.
Solution
Since λ1 and λ2 are distinct eigenvalues it follows that Eλ1 ∩ Eλ2 = {0}, so
dim(Eλ1 + Eλ2 ) = dim(Eλ1 ) + dim(Eλ2 ) − dim(Eλ1 ∩ Eλ2 )
= dim(Eλ1 ) + dim(Eλ2 )
≥ n since dim(Eλ1 ) = n − 1 and dim(Eλ2 ) ≥ 1

Therefore F n = Eλ1 + Eλ2 and hence F n = Eλ1 ⊕ Eλ2 since we also have Eλ1 ∩ Eλ2 = {0}. It follows
that if β1 and β2 are bases for Eλ1 and Eλ2 respectively, then β = β1 ∪ β2 is a basis for F n consisting
of eigenvectors. Thus A is diagonalizable.

Question 6

This question was not marked.

It is still important to work through the solutions and compare them with your own attempts.
  
a b
Let T : M2×2 (R) → M2×2 (R) denote the projection on W = : a − d = b − c = 0 along
   c d
a b
U = span :a+d=b+c=0 .
c d

8
 MAT3701/201

       
1 0 0 1 0 0 0 0
(a) Find the matrix representation of T with respect to β = , , , .
0 0 0 0 1 0 0 1

(b) Find the formula for T.

Solution

(a) An element in W is of the form  

a b
b a

and an element in U is of the form  

c d
−d −c

Now      
1 0 1 1 0 1 1 0
= +2
0 0 2 0 1 0 −1
so            
1 0 1 1 0 1 1 0 0 1 0 0 1 0 0
T =2 =2 +0 +0 +2
0 0 0 1 0 0 0 0 1 0 0 1

Similarly
           
0 1 1 0 1 1 0 1 0 1 1 0 0 0 0
T =2 =0 +2 +2 +0
0 0 1 0 0 0 0 0 1 0 0 1

           
0 0 1 0 1 1 0 1 0 1 1 0 0 0 0
T =2 =0 +2 +2 +0
1 0 1 0 0 0 0 0 1 0 0 1
           
0 0 1 1 0 1 1 0 0 1 0 0 1 0 0
T =2 =2 +0 +0 +2
0 1 0 1 0 0 0 0 1 0 0 1

Thus  
1 0 0 1
 0 1 1 0 
[T ]β = 21 
 0

1 1 0 
1 0 0 1

(b) Since          
a b 1 0 0 1 0 0 0 0
=a +b +c +d ,
c d 0 0 0 0 1 0 0 1

9
 we have
     
a b 1 0 0 1
T = aT + bT
c d 0 0 0 0
   
0 0 0 0
+cT + dT
1 0 0 1
 1   1
  1
  1 
0 0 0 0
= a 2 1 +b 1 2 +c 1 2 +d 2 1
0 2 2
0 2
0 0 2
" #
a+d b+c
2 2
= b+c a+d
2 2

Question 7

Let T : M4×4 (C) → M4×4 (C) be the linear operator over C defined by
 
0 0 0 0
 1 0 0 0 
T (X) = AX, where A =   0 1 1 0 ,


0 0 0 −1
and let W be the T –cyclic subspace of M4×4 (C) generated by A.
(a) Find the T –cyclic basis for W. (17)
(b) Find the characteristic polynomial of TW . (5)
[22]

Solution

(a)
   
0 0 0 0 0 0 0 0
1 0 0 0 0 0 0 0
A= 0
, T (A) =  ,
1 1 0 1 1 1 0
0 0 0 -1 0 0 0 1
   
0 0 0 0 0 0 0 0
0 0 0 0 3
 0 0 0 0
T 2 (A) = 
1 , T (A) =  
1 1 0 1 1 1 0
0 0 0 -1 0 0 0 1

Since A, T (A) and T 2 (A) are linearly independent and T 3 (A) = T (A), it follows that the
T –cyclic basis for W is {A, T (A) , T 2 (A)} .
(b) Since T 3 (A) = T (A) from (a), we have cTW = (−1)3 (t3 − t) = −(t3 − t).

10
 MAT3701/201

Question 8
 
0 0 1
Let A =  1 0.5 0  .
0 0.5 0

(a) Show that A is a regular transition matrix. (13)

(b) Find the smallest disc centered at the origin of the complex plane in which the eigenvalues of A
lie. (4)
(c) Find an eigenvalue of A on the circumference of your disc in (b), and a corresponding eigenvalue.
(13)
(d) Find lim Am . (5)
m→∞

[35]

Solution

(a) A is a transition matrix since its entries are nonnegative and the column sums are all equal to
1. It is also regular since the entries of
 1   1  1 1 1
0 2 0 0 2 0 4 8 2
4 1 1  1 1  5 9 1
A =  2 4 1  2 4 1 =  8 16 4 
1 1 1 1 1 5 1
2 4
0 2 4
0 8 16 4

are all positive.

(b) It is the disc with radius min{ρ(A), ν(A)} =min{1.5, 1} = 1.
(c) Since A is a regular transition matrix, λ = 1 is always an eigenvalue. Now

        
x1 x1 −1 0 1 x1 0
A x2  = x2  ⇔  1 −0.5 0  x2  = 0
x3 x3 0 0.5 −1 x3 0
  1
x1 4
⇔ x2  = t  12  , t ∈ R (show)
  
x3 1
4

(d)  
1 1 1
4 4 4
lim Am =  1 1 1
.
 
m→∞ 2 2 2
1 1 1
4 4 4

Total [100]

11
 MAT3701/202/2/2016

Tutorial Letter 202/2/2016

LINEAR ALGEBRA

MAT3701

Semester 2

Department of Mathematical Sciences

This tutorial letter contains

solutions to Assignment 02.

BAR CODE

university
Learn without limits. of south africa
 ASSIGNMENT 02
Solution
UNIQUE ASSIGNMENT NUMBER: 708742

Please note that we will only mark a selection of the questions. It is therefore in your own best
interest to attempt all the questions. The fact that a question is not marked does not mean that it
is less important than one that is marked. We try to cover the whole syllabus over the two semesters
(4 assignments) and to use the assignments to help you prepare for the exam. It is therefore good
practice to work through a complete set of four assignments for a given year, and for this reason the
assignments and worked solutions of previous years are made available on myUnisa under Additional
Resources – see also the letter MAT3701 Exam Preparation under Additional Resources.

Worked solutions for this assignment will be made available on myUnisa shortly after the due date.
Your answers to the assignment questions should be fully motivated.

For this assignment, Questions 2, 5, 7 and 8 will be marked.

Question 1

This question was not marked.

It is still important to work through the solutions and compare them with your own attempts.
 
2 2 ∗ 2 i
Define <, >: C × C → C by < x, y >= y Ax where A = and x and y are column vectors
−i 2
in C 2 . Show that <, > is an inner product on C 2.

Solution
     
x1 y1 z1
Let x = , y= and z = be column vectors in C 2 and let c ∈ C.
x2 y2 z2
IP1

hx + z, yi = y ∗ A (x + z)
= y ∗ Ax + y ∗ Az
= hx, yi + hz, yi

Thus IP1 is satisfied.

IP2

hcx, yi = y ∗ A (cx)
= c(y ∗ Ax)
= chx, yi

2
 MAT3701/202

Thus IP2 is satisfied.

IP3

hx, yi = y ∗ Ax
= ȳ t Ax
= y t Āx̄
= (y t Āx̄)t
= x̄t Āt (y t )t
= x∗ A ∗ y
= x∗ Ay since A = A∗
= hy, xi

Thus IP3 is satisfied.

IP4
  
2 i x1
hx, xi = (x1 , x2 )
−i 2 x2
= 2x1 x1 + ix1 x2 − ix2 x1 + 2x2 x2
= (x1 − ix2 )(x1 + ix2 ) + x1 x1 + x2 x2
= (x1 + ix2 )(x1 + ix2 ) + x1 x1 + x2 x2
= |x1 + ix2 |2 + |x1 |2 + |x2 |2 ≥ 0

and

hx, xi = 0 ⇔ |x1 + ix2 |2 = |x1 |2 = |x2 |2 = 0

⇔ |x1 + ix2 | = |x1 | = |x2 | = 0
⇔ x1 + ix2 = x1 = x2 = 0
⇔ x1 = x2 = 0
⇔x=0

Thus IP4 is satisfied. Since all four conditions are satisfied it follows that <, > is an inner product on
C 2.

Question 2

Given that the system

x1 + x 2 + x3 = 0
x1 − x2 = 6
2x2 + x3 = 0

is inconsistent, find a least squares approximate solution in R3 (see SG: Section 12.6). [10]

3
Solution

Use the method described in Section 12.6 of the Study Guide. Denote the coefficient matrix and
constant column of the system by
   
1 1 1 0
A= 1
 −1 0 and y = 6
 
0 2 1 0

respectively. Then    
2 0 1 6
∗ ∗
A A = 0 6 3 and A y = −6
  
1 3 2 0
The least squares approximate solution x0 is obtained from A∗ Ax0 = A∗ y, which yields
 
3
x0 = −1 (show)
0

Question 3

This question was not marked.

It is still important to work through the solutions and compare them with your own attempts.

(a) (F4: Exercise 6.4.4, p 375) Let T and U be self-adjoint operators on an inner product space V.
Prove that T U is self-adjoint if and only if T U = U T.

(b) The following exercise illustrates Theorem 6.15. Let T : C 3 → C 3 be the linear operator with
matrix representation  
−1 i 0
A= i −1 0 
0 0 −1 + i
with respect to the standard basis.

(i) Show that T is normal.

(ii) Find T (1, 1, 1) and T ∗ (1, 1, 1) and verify that kT (1, 1, 1)k = kT ∗ (1, 1, 1)k.
(iii) Show that v = (1, 1, 1) is an eigenvector of T corresponding to λ = −1 + i. Verify that
T ∗ (v) = λv.
(iv) Find an eigenvector u of T corresponding to −1 − i. Verify that hu, vi = 0 with v as in (c).

Solution

4
 MAT3701/202

(a)

T U is self-adjoint ⇔ T U = (T U )∗
⇔ T U = U ∗ T ∗ property of the adjoint
⇔ T U = U T since T and U are self-adjoint
  
−1 i 0 −1 −i 0
(b) (i) AA∗ =  i −1 0   −i −1 0 
 0 0 −1 + i 0 0 −1 − i
2 0 0
=  0 2 0 
0 0 2
= 2I3
SimilarlyA∗ A = 2I3 . So
AA∗ = A∗ A
which can be written as
[T ]β [T ]∗β = [T ]∗β [T ]β

where β is the standard basis for C 3 . Since β is orthonormal with respect to the standard
inner product, we have
[T ]β [T ∗ ]β = [T ∗ ]β [T ]β
therefore
[T T ∗ ]β = [T ∗ T ]β
which implies that
T T ∗ = T ∗T
so T is normal.
(ii)

[T (1, 1, 1)]β = [T ]β (1, 1, 1)β

 
1
= A 1 

1
 
−1 + i
=  −1 + i 
−1 + i

So
T (1, 1, 1) = (−1 + i) (1, 1, 1) ......(1)
and
kT (1, 1, 1)k2 = k(1 + i)(1, 1, 1)k2 = |1 + i|2 k(1, 1, 1)k2 = 2 · 3 = 6

5
 Similarly

[T ∗ (1, 1, 1)]β = [T ∗ ]β [(1, 1, 1)]β

 
1
∗
= A 1 
1
 
−1 − i
=  −1 − i 
−1 − i

so
T ∗ (1, 1, 1) = (−1 − i) (1, 1, 1) ......(2)
and

kT ∗ (1, 1, 1)k2 = 6
Thus
kT (1, 1, 1)k = kT ∗ (1, 1, 1)k
(iii) It follows from (1) that v = (1, 1, 1) is an eigenvector of T corresponding to λ = −1 + i and
form (2) that v is an eigenvector of T ∗ corresponding to λ.
(iv) Calculate E−1−i (A). In matrix form, the associated homogeneous system is
   
−1 − (−1 − i) i 0 i i 0
 i −1 − (−1 − i) 0  =  i i 0 
0 0 (−1 + i) − (−1 − i) 0 0 2i

and therefore
  
 t 
E−1−i (A) =  −t : t ∈ C

0
 
 
 1 
= span  −1 
0
 

So
E−1−i (T ) = span {(1, −1, 0)}
since A is the matrix representation of T with respect to the standard basis. Thus u =
(1, −1, 0) is an eigenvector of T corresponding to −1 − i and

hu, vi = h(1, −1, 0) , (1, 1, 1)i

= 0

as required.

6
 MAT3701/202

Question 4

This question was not marked.

It is still important to work through the solutions and compare them with your own attempts.

Let U be a unitary operator on a finite-dimensional inner product space V and let W be a U -invariant
subspace of V. Prove that

(a) U (W ) = W ;

(b) W ⊥ is U -invariant.

Solution

(a) Since W is U -invariant, U (W ) ⊆ W . Thus U maps W into W . Since U is unitary it is one-to-one

and since W is finite dimensional, U maps W onto W, so U (W ) = W .

(b)

hU (W ⊥ ), W i = hU (W ⊥ ), U (W )i from (a)
= hW ⊥ , U ∗ U (W )i property of the adjoint
= hW ⊥ , W i since U ∗ U = IV
=0

It follows that U (W ⊥ ) ⊆ W ⊥, so W ⊥ is U -invariant.

Question 5

Let f : R2 → R2 be defined by
f (a, b) = (1 + b, 2 + a)

(a) Prove that f is a rigid motion. (6)

(b) Express f in the form f = g ◦ T where g is a translation and T an orthogonal operator on R2 .

Find the vector v0 by which g translates. If T is a rotation, find the angle θ through which it
rotates; and if T is a reflection about a line L through the origin, find the equation of L and the
angle α it makes with the positive x–axis. (10)
[16]

Solution

7
 (a) For all (a, b) , (c, d) ∈ R2

f (a, b) − f (c, d) = (1 + b, 2 + a) − (1 + d, 2 + c) = (b − d, a − c)

so

kf (a, b) − f (c, d) k2 = k (b − d, a − c) k2
= (b − d)2 + (a − c)2
= (a − c)2 + (b − d)2
= k (a, b) − (c, d) k2

hence f is a rigid motion.

(b) Since

f (a, b) = (1, 2) + (b, a)

= (g ◦ T ) (a, b)

it follows that
g (a, b) = (1, 2) + (a, b)

Thus g translates by (1, 2) = f (0). T is defined by

T (a, b) = (b, a)

Let β be the standard ordered basis for R2 , then

 
0 1
[T ]β =
1 0

T is orthogonal since
[T ]∗β [T ]β = [T ]β [T ]∗β = I2.

Since  
det [T ]β = −1

and

cos π2 sin π2
   
0 1
[T ]β = =
1 0 sin π2 cos π2
it follows from Friedberg: Example 5, page 383 that T is a reflection in the line L making an
angle α = π4 with the positive x-axis. Thus the equation of L is y = tan( π4 ) = x.

8
 MAT3701/202

Question 6

This question was not marked.

It is still important to work through the solutions and compare them with your own attempts.

Let V = M2×2 (C) denote the inner product space over C with respect to the Frobenius inner product,
i.e.    
z1 z2 y1 y2
, = z1 ȳ1 + z2 ȳ2 + z3 ȳ3 + z4 ȳ4
z3 z4 y3 y4
Let T : V → V denote the orthogonal projection on
    
1 1 i 1 −1 i
W = span (S) where S = ,
2 i 1 2 i −1

(a) Show that S is an orthonormal subset of V.

 
z1 z2
(b) Find a formula for T expressed as a single matrix in terms of z1 , z2 , z3 and z4 .
z3 z4
 
1 1
(c) Use T to express A = as A = B + C where B ∈ W and C ∈ W ⊥ .
1 1
(d) Find the matrix in W closest to A.

Solution
   
1 1 i 1 −1 i
Let v1 = and v2 = .
2 i 1 2 i −1

(a)
   
1 1 i 1 i
hv1 , v1 i = ,
4 i 1 i 1
1
= (1 · 1̄ + i · ı̄ + i · ı̄ + 1 · 1̄)
4
1
= (1 − i2 − i2 + 1)
4
= 1

Similarly hv2 , v2 i = 1.
   
1 1 i −1 i
hv1 , v2 i = ,
4 i 1 i −1
1

= 1 · −1 + i · ı̄ + i · ı̄ + 1 · −1
4
1
= (−1 − i2 − i2 − 1) = 0
4

Thus S is an orthonormal subset of V .

9
 (b)
      
z1 z2 z1 z2 z1 z2
T = , v1 v1 + , v2 v2
z3 z4 z3 z4 z3 z4
   
1 1 i 1 −1 i
= (z1 − iz2 − iz3 + z4 ) + (−z1 − iz2 − iz3 − z4 )
4 i 1 4 i −1
 
1 z1 + z4 z2 + z3
=
2 z2 + z3 z1 + z4

(c)
A = T (A) + (I − T )(A) where T (A) ∈ W and (I − T )(A) ∈ W ⊥
   
1 1+1 1+1 1 1−1 1−1
= +
2 1+1 1+1 2 1−1 1−1

Thus B = A and C = O2 .
(d) Since A ∈ W it follows that the matrix closest to A is A itself.

Question 7

It is given that A ∈ M3×3 (C) is a normal matrix with eigenvalues 1 and i and corresponding
eigenspaces  
1 1
E1 = span (2, 2, 1), (1, −2, 2)
3 3
and  
1
Ei = span (−2, 1, 2)
3
(a) Find the spectral decomposition of A. (10)
(b) Find A. (1)
[11]

Solution
(a) 2 1
 5 2 4

3 3 2 9 9 9
2 1

   
2  3
−2  
3 3  −2 

P1 =   = 2 8
3 3 
−2
9 9 9 
  1 2  
3 3 3
1 2 4 −2 5
3 3 9 9 9
and  −2  4 −2 −4
 
3 9 9 9
   
 1
  −2 1 2
  −2 1 2

Pi = 
 3

 3 3 3
=
9 9 9


   
2 −4 2 4
3 9 9 9

10
 MAT3701/202

The spectral decomposition is

A = 1 · P 1 + i · Pi
−2 −4
5 2 4
  4

9 9 9 9 9 9
   
−2 
2 8
  −2 1 2

=1·
9 9 9 
+i·
9 9 9


   
4 −2 5 −4 2 4
9 9 9 9 9 9

(b)  5+4i 2−2i 4−4i


9 9 9
 
−2+2i 
 2−2i 8+i

A=
 9 9 9 
 
4−4i −2+2i 5+4i
9 9 9

Question 8
 
1 −3
Let A = and suppose k·k denotes the Euclidean norm.
3 −1
(a) Find kAk , kA−1 k and cond(A). (7)
e such that Ax = b, kbk = 1 and kb − Ae
(b) Suppose that we have vectors x and x xk ≤ 0.001. Use (a)
−1
to determine upper bounds for kex − A bk (the absolute error) and ke x − A−1 bk / kA−1 bk (the
relative error). (6)
[13]
Solution
(a)    
∗1 3 1 −3 10 −6
A A= =
−3 −1 3 −1 −6 10
thus
λ − 10 6
det(λI − A∗ A) =
6 λ − 10
= (λ − 10)2 − 62
= (λ − 16)(λ − 4)
so
√ 1 1
kAk = 16 = 4, A−1 = √ = and cond (A) = kAk · A−1 = 2
4 2
(b)
1
e − A−1 b = A−1 (Ae
x x − b) ≤ A−1 kAe
x − bk ≤ (0.001) = 0.0005
2
and
x − A−1 bk
ke ke
x − xk kAe
x − bk
= ≤ cond (A) ≤ 2(0.001) = 0.002
kA−1 bk kxk kbk
Total [50]

11