MAT3701/201/1/2015

Tutorial Letter 201/1/2015

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
 ONLY FOR SEMESTER 1 STUDENTS
ASSIGNMENT 01
Based on Study Units 1 - 9
FIXED CLOSING DATE: 9 MARCH 2015
UNIQUE NUMBER: 601473

Please note that we will only mark a selection of the questions. It is therefore in your own inter-
est 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 under Addi-
tional Resources on myUnisa – see also the letter MAT3701 Exam Preparation under Additional
Resources.

Worked solutions to all the questions for this assignment 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 2, 5, 6, and 8 will be marked.

QUESTION 1

Given that
B= 5 3t + 2t2 ; 2 + t2 ; 1 + 2t; 2t + 7t2 ; 2 + t + 10t2
is a generating set for P2 (C) ; find a basis for P2 (C) from among the vectors of B:

SOLUTION

Write the coefficients of the polynomials in B as the columns of a matrix and reduce to echelon
form:
2 3 2 3
5 2 1 0 2 1 4 1 14 18 R1 2R3
4 3 0 2 2 1 ! 5 4 1 1 2 9 11 5 R2 + R3
2 1 0 7 10 2 1 0 7 10
2 3 2 3
0 3 1 5 7 R1 + R2 1 1 2 9 11 R2
! 4 1 1 2 9 11 5 4
! 0 3 1 5 7 5 R1
0 3 4 25 32 R3 + 2R2 0 0 5 20 25 R3 + R1

Since the leading coefficients occur in columns 1; 2 and 3; it follows that the first three polynomials
in B is a basis for P2 (C) ; namely,

5 3t + 2t2 ; 2 + t2 ; 1 + 2t :

2
 MAT3701/201

QUESTION 2

Let f0 (x) ; f1 (x) ; f2 (x) denote the Lagrange polynomials over R associated with 0; 1 and 2;
respectively.

(a) Calculate f0 (x) ; f1 (x) and f2 (x) :

(b) Use the Lagrange interpolation formula to express, 1; x; and x2 as linear combinations of
f0 (x) ; f1 (x) ; and f2 (x) :

(c) Without any further calculations, explain why

= ff0 ; f1 ; f2 g

is a basis for P2 (R) :

(d) Let
= 1; x; x2

and write down the change of coordinate matrix Q which changes coordinates into –
coordinates.

(e) Use Q to express

p (x) = 1 + x + x2

as a linear combination of :

(f) Check your answer in (e) by using the Lagrange interpolation formula to express p (x) as a
linear combination of :

SOLUTION

(a)

(x 1) (x 2) 1
f0 (x) = = (x 1) (x 2)
(0 1) (0 2) 2
(x 0) (x 2)
f1 (x) = = x (x 2)
(1 0) (1 2)
(x 0) (x 1) 1
f2 (x) = = x (x 1)
(2 0) (2 1) 2

(b)

1 = 1 f0 (x) + 1 f1 (x) + 1 f2 (x)

x = 0 f0 (x) + 1 f1 (x) + 2 f2 (x)
x2 = 0 f0 (x) + 1 f1 (x) + 4 f2 (x)

3
 (c) It follows from (b) that generates P2 (R), and since j j = 3 = dim (P2 (R)) ; it also is a basis
for P2 (R).
(d) It follows from (b) that 2 3
1 0 0
4
Q= 1 1 1 5:
1 2 4
(e) 2 32 3 2 3
1 0 0 1 1
[p (x)] = Q [p (x)] = 4 1 1 1 54 1 5 = 4 3 5;
1 2 4 1 7
so
p (x) = 1 f0 (x) + 3 f1 (x) + 7 f2 (x) :
(f) According to the Lagrange interpolation formula,
p (x) = p (0) f0 (x) + p (1) f1 (x) + p (2) f2 (x)
= 1 f0 (x) + 3 f1 (x) + 7 f2 (x)

QUESTION 3

1 i
Let A = ; and let T : M2 2 (C) ! M2 2 (C) be defined by T (X) = XA:
i 1
(a) Show that T is a linear transformation over C:
(b) Find a basis for R (T ) :
(c) Find a basis for N (T ) :
(d) Determine whether or not M2 2 (C) = R (T ) N (T ) :

SOLUTION

(a) For X; Y 2 M2 2 (C) and z 2 C;

T (X + Y ) = (X + Y ) A
= XA + Y A
= T (X) + T (Y )

and
T (zX) = (zX) A
= z (XA)
= zT (X) :

Thus, T is a linear transformation since it satisfies both conditions for a linear transformation.

4
 MAT3701/201

(b) Since
1 0 0 1 0 0 0 0
= A11 = ; A12 = ; A21 = ; A22 =
0 0 0 0 1 0 0 1

is a basis for M2 2 (C) ;

R (T ) = span fT (A11 ) ; T (A12 ) ; T (A21 ) ; T (A22 )g
= span fA11 A; A12 A; A21 A; A22 Ag
1 i i 1 0 0 0 0
= span ; ; ;
0 0 0 0 1 i i 1
1 i 0 0
= span ;
0 0 1 i

since
1 i i 1 0 0 0 0
i = and i = :
0 0 0 0 1 i i 1
Thus,
1 i 0 0
= ;
0 0 1 i

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

z1 z2
(c) Let X = ; where z1 ; z2 ; z3 ; and z4 are complex numbers.
z3 z4
X 2 N (T ) , T (X) = 0
, XA = 0
z1 z2 1 i 0 0
, =
z3 z4 i 1 0 0
z1 + iz2 iz1 z2 0 0
, =
z3 + iz4 iz3 z4 0 0
z1 + iz2 = 0 z3 + iz4 = 0
, and
iz1 z2 = 0 iz3 z4 = 0
, z1 + iz2 = 0 and z3 + iz4 = 0;

since i (z1 + iz2 ) = iz1 z2 and i (z3 + iz4 ) = iz3 z4 :

, z2 = iz1 and z4 = iz3 :

Thus,
z1 iz1 1 i 0 0
X= = z1 + z3 :
z3 iz3 0 0 1 i

A basis for N (T ) is therefore

1 i 0 0
= ; :
0 0 1 i

5
 (d) Since = ; it follows that [ cannot be a basis for M2 2 (C) ; and therefore M2 2 (C) is
not the direct sum of R (T ) and N (T ) :

QUESTION 4

Let V be a vector space over F with basis = fv1 ; :::; vn g, and let T : V ! V be a linear operator.

(a) Show that T 2 = IV , [T ]2 = In :

(b) Let V = M2 2 (C) ; considered as a vector space over C; and let T : M2 2 (C) ! M2 2 (C) be
the linear operator defined by

1 i i
T (X) = AX; where A = :
1 i i

Determine whether or not T is diagonalizable. If it is, find a basis for M2 2 (C) such that
[T ] is diagonal, and write down [T ] :

SOLUTION

(a) T 2 = IV , [T 2 ] = [IV ] , [T ]2 = In ; since [T 2 ] = [T ]2 (F4: Theorem 2.11) and

[IV ] = In (F4: Theorem 2.12(d)).

(b) Let
1 0 0 1 0 0 0 0
= A11 = ; A12 = ; A21 = ; A22 = :
0 0 0 0 1 0 0 1

Then
1 i 0
T (A11 ) = AA11 = = (1 i) A11 + 0 A12 + (1 i) A21 + 0 A22
1 i 0
0 1 i
T (A12 ) = AA12 = = 0 A11 + (1 i) A12 + 0 A21 + (1 i) A22
0 1 i
i 0
T (A21 ) = AA21 = = i A11 + 0 A12 + i A21 + 0 A22
i 0
0 i
T (A22 ) = AA22 = = 0 A11 + i A12 + 0 A21 + i A22
0 i

so that 2 3
1 i 0 i 0
6 0 1 i 0 i 7
[T ] = 6
4 1
7
i 0 i 0 5
0 1 i 0 i

6
 MAT3701/201

and
1 i 0 i 0
0 1 i 0 i
det [T ] I =
1 i 0 i 0
0 1 i 0 i
1
0 ii 0 1 i i
= (1 i 0 i) 0 +i 1 i 0 0
1 i 0 i 0 1 i i
(expanding along the first row)
1 i i 1 i i
= (1 i ) (i ) i (1 i)
1 i i 1 i i
(expanding along the second row of both)
1 i i
= [(1 i ) (i ) i (1 i)]
1 i i
= [(1 i ) (i ) i (1 i)]2
2 2
=
= 2
( 1)2 :

Therefore the eigenvalues of T are 0 and 1, both with multiplicity 2:

2 3 2 3
1 i 0 i 0 1 i 0 i 0
6 0 1 i 0 i 7 6 0 1 i 0 i 7
[T ] 0I = 6
4 1
7!6
5 4
7;
5 (:::i)
i 0 i 0 0 0 0 0
0 1 i 0 i 0 0 0 0

so
n rank [T ] 0I = 4 2 = 2 = algebraic multiplicity of 0;

and 2 3 2 3
i 0 i 0 1 0 1 0
6 0 i 0 7
i 7 6 0 1 0 1 7
[T ] I=6
4 1 5 !6
4
7 (:::ii)
i 0 i 1 0 0 0 0 0 5
0 1 i 0 i 1 0 0 0 0

so that
n rank [T ] I =4 2 = 2 = algebraic multiplicity of 1:

Since the geometric and algebraic multiplicities are equal for both eigenvalues, it follows that
T is diagonalizable.
From (i), 82 3 2 39
>
> i 0 >
>
<6 7 6 7=
0 i
E0 [T ] = span 64
7; 6 7
>
> 1+i 5 4 0 5>
>
: ;
0 1+i

7
 so that
i 0 0 i
E0 (T ) = span ;
1+i 0 0 1+i

with basis
i 0 0 i
1 = ; ;
1+i 0 0 1+i

and, from (ii), 82 3 2 39

>
> 1 0 >
>
<6 7 6 7=
0 7 6 1
E1 [T ] = span 64 ; 7
>
> 1 5 4 0 5>
>
: ;
0 1

so that
1 0 0 1
E1 (T ) = span ;
1 0 0 1

with basis
1 0 0 1
2 = ; :
1 0 0 1

Therefore
i 0 0 i 1 0 0 1
= 1 [ 2 = ; ; ;
1+i 0 0 1+i 1 0 0 1

is a diagonalizing basis such that

2 3
0 0 0 0
6 0 0 0 0 7
[T ] = 6
4 0
7:
0 1 0 5
0 0 0 1

QUESTION 5

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

(a) Show that T 2 = 0 , R (T ) N (T ).

(b) Suppose
82 that
3 V
2 =39C 4 and F = C: Find a formula for T such that T 2 = 0; N (T ) =
>
> 1 0 >>
<6 7 6 0 7=
1 7
span 64 0 5;
6 7 , and rank (T ) = 2:
4 1 5>
>
> >
: ;
0 0

8
 MAT3701/201

SOLUTION

(a)
T2 = 0 , T 2 (v) = 0 for all v 2 V
, T (T (v)) = 0 for all v 2 V
, T (v) 2 N (T ) for all v 2 V
, R (T ) N (T ) :

(b) From (a), R (T ) N (T ) ; and since they are both of dimension two, it follows that R (T ) =
N (T ) :
82 3 2 3 2 3 2 39
>
> 1 1 0 0 >
<6 7 6 7 6 7 6 7> =
0 7 6 1 7 6 0 7 6 0 7
Since = 6 ; ; ; 4
4 0 5 4 0 5 4 1 5 4 0 5> is a basis for C ; we may choose, for example,
>
> >
: ;
0 0 0 1
2 3 2 3 2 3 2 3 2 3 2 3 2 3 2 3
1 1 1 0 0 0 0 0
6 0 7 6 1 7 6 1 7 6 0 7 6 0 7 6 0 7 6 0 7 6 0 7
T6 7 6 7 6 7 6 7 6 7 6 7 6 7 6 7
4 0 5 = 4 0 5 ; T 4 0 5 = 4 0 5 ; T = 4 1 5 = 4 0 5 ; and T 4 0 5 = 4 1 5 ;
0 0 0 0 0 0 1 0

then 82 3 2 39
>
> 1 0 >
>
<6 7 6 7=
1 7; 6 0
R (T ) = N (T ) = span 64
7 :
>
> 0 5 4 1 5>
>
: ;
0 0
Since 2 3 2 3 2 3 2 3
0 1 1 1
6 1 7 6 1 7 6 0 7 6 1 7
T6 7
4 0 5=T
6 7
4 0 5 T6 7 6
4 0 5=4
7;
0 5
0 0 0 0
we have
2 3 2 3 2 3 2 3 2 3
a 1 0 0 0
6 b 7 6 0 7 6 1 7 6 7 6 0 7
T6 7 6 7 6 7 + cT 6 0 7 + dT 6 7
4 c 5 = aT 4 0 5 + bT 4 0 5 4 1 5 4 0 5
d 0 0 0 1
2 3 2 3 2 3 2 3
1 1 0 0
6 1 7 6 1 77 6 7 6 7
= a6 7 + b6 + c 6 0 7 + d6 0 7
4 0 5 4 0 5 4 0 5 4 1 5
0 0 0 0
2 3
a b
6 a b 7
= 6
4 d 5:
7

9
QUESTION 6 (Friedberg: Exercise 5.1.11, p. 258.)

A scalar matrix is a square matrix of the form I for some scalar ; that is, a scalar matrix is a
diagonal matrix in which all the diagonal entries are equal.

(a) Prove that if a square matrix A is similar to a scalar matrix I, then A = I:

(b) Show that a diagonalizable matrix having only one eigenvalue is a scalar matrix.

SOLUTION

(a) If A is similar to a scalar matrix I, then

A = X 1 ( I) X; for some nonsingular matrix X

= X 1 IX
= X 1X
= I:

(b) If A is diagonalizable, then A = X 1 DX for some nonsingular matrix X and diagonal matrix D
which contains the eigenvalues of A. But the eigenvalues of A are all equal, so D = I; where
denotes the (only) eigenvalue of A. It follows from (a) that A = I, a scalar matrix.

QUESTION 7

1
(a) If A 2 Mn n (R) is a symmetric regular matrix, show that lim Am = E; where each entry of
m!1 n
E 2 Mn n(R) is equal to 1:
2 3
0:5 0 1
(b) Let A = 4 0:5 0 0 5 :
0 1 0

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

(ii) Find lim Am :
m!1

10
 MAT3701/201

SOLUTION

(a) Since A is regular, it follows from F4: Theorem 5.20 (e) that lim Am = L; where the columns
m!1
of L are all equal to a unique probability vector v: This means that the entries in each row of
L are equal. Since A is also symmetric, L is symmetric as well, since
T m
LT = lim Am = lim (Am )T = lim AT = lim Am = L:
m!1 m!1 m!1 m!1

Therefore, the entries in v are all equal. Since we also know that the entries in v are nonneg-
ative and adds up to 1; it follows that
2 3
1
16 . 7
v = 4 .. 5 :
n
1

Thus, L = n1 E:
(b) (i) All the entries of A are nonnegative and each column adds up to one, thus A is a
transition matrix. Further,
2 1 3 2 9 1 5 3
4
1 12 16 4 8
A2 = 4 14 0 12 5 and A4 = 4 16 5 1
4
1 5
8
:
1 1 1 1
2
0 0 8 2 4

Since all the entries in A4 are positive, it follows that A is regular as well.
(ii) We have to find the eigenvector v of A corresponding to = 1 that is also a probability
vector: 2 1 3 2 3
2
0 1 1 0 2 2R1
A I= 4 1
1 5
0 ! 0 4 1 1 5 R2 + R1
2
0 1 1 0 1 1
2 3
2
4 1 5 is a solution to this system, and therefore
1
2 3 2 3
2 2 2 2
14 5 14
v= 1 and lim A = m
1 1 1 5:
4 m!1 4
1 1 1 1

QUESTION 8

Let T : M4 4 (C) ! M4 4 (C) be the linear operator over C defined by T (X) = AX; where
2 3
0 0 0 0
6 1 0 1 0 7
A=6 4 0 1 0 0 5;
7

0 0 0 1
and let W be the T –cyclic subspace of M4 4 (C) generated by A:

11
 (a) Find the T –cyclic basis for W: (7)

(b) Find the characteristic polynomial of TW : (2)

[9]

Total [50]

SOLUTION

(a)
2 3 2 3
0 0 00 0 0 0 0
6 1 0 1 7
0 7 6 0 1 0 0 7
A = 6
4 0 5 ; T (A) = A2 = 6
4
7;
1 00 1 0 1 0 5
0 0 01 0 0 0 1
2 3
0 0 0 0
6 1 0 1 0 7
T 2 (A) = T (T (A)) = T A2 = A3 = 6 4 0 1 0
7 = A:
0 5
0 0 0 1

Since A and T (A) are linearly independent and T 2 (A) = A, it follows that the T –cyclic basis
for W is fA; T (A)g = fA; A2 g :

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

12
 MAT3701/202/1/2015

Tutorial Letter 202/1/2015

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
 ONLY FOR SEMESTER 1 STUDENTS
ASSIGNMENT 02
Based on Study Units 10 - 17
FIXED CLOSING DATE: 7 APRIL 2015
UNIQUE NUMBER: 601488

Please note that we will only mark a selection of the questions. It is therefore in your own inter-
est 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 under Addi-
tional Resources on myUnisa – see also the letter MAT3701 Exam Preparation under Additional
Resources.

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

For this assignment, question 3, 6, 7 and 8 will be marked.

QUESTION 1 (Friedberg: Exercise 6.1.22, p. 339.)

Let V be a real or complex vector space (possibly infinite–dimensional), and let be a basis for
V: For x; y 2 V there exist v1 ; :::; vn 2 such that
X
n X
n
x= ai vi and y = bi vi :
i=1 i=1

Define
X
n
hx; yi = ai b i :
i=1

(a) Prove that h ; i is an inner product on V and that is an orthonormal basis for V: Thus every
real or complex vector space may be regarded as an inner product space.

(b) Prove that if V = Rn or V = C n and is the standard ordered basis, then the inner product
defined above is the standard inner product.

SOLUTION

Let x; y; z 2 V . Since each of these vectors can be expressed as a finite linear combination of
vectors in , there exists an integer n such that
X
n X
n X
n
x= ai vi ; y = bi v i ; and z = ci vi ;
i=1 i=1 i=1

2
 MAT3701/202

where
ai ; bi ; ci 2 F and v1 ; v2 ; : : : vn 2 :
(It may be necessary to set some of the scalars ai , bi , ci 2 F equal to 0 in order to find a common
integer n.)

(a) We need to verify conditions (a) - (d) in the definition of an inner product.
Condition (a)
* n +
X X
n
hx + y; zi = (ai + bi )vi ; ci vi
i=1 i=1
X
n
= (ai + bi )ci ; by the definition above
i=1
Xn X
n
= ai c i + bi c i
i=1 i=1
= hx; zi + hy; zi :

Condition (b)
Let c 2 F . Then
* +
X
n X
n
hcx; yi = c ai v i ; bi vi
* ni=1 i=1
+
X Xn
= (cai )vi ; bi v i
i=1 i=1
X
n
= (cai )bi
i=1
Xn
= c(ai bi )
i=1
Xn
= c ai b i
i=1
= c hx; yi :

3
 Condition (c)
* n +
X X
n
hy; xi = bi vi ; ai vi
i=1 i=1
X
n
= bi ai
i=1
Xn
= bi ai
i=1
X
n
= bi ai
i=1
X
= ai b i
= hx; yi :

Condition (d)
* n +
X X
n
hx; xi = ai vi ; ai vi
i=1 i=1
X
n
= ai ai
i=1
Xn
= jai j2
i=1
> 0 if x 6= 0; since ai 6= 0 for some i:
2 3 2 3
a1 b1
6 7 6 7
(b) If x = 4 ... 5, y = 4 ... 5 2 V; where V = Rn or V = C n ; and = fv1 ; : : : ; vn g is the
an bn
standard ordered basis, then
X
n X
n
x= ai vi and y = bi v i ;
i=1 i=1

so that 2 3 2 3
* a1 b1 + *X +
6 .. 7 ; 6 .. 7 =
n Xn Xn
hx; yi = 4 . 5 4 . 5 ai vi ; bi vi = ai b i ;
an bn i=1 i=1 i=1

which is the standard inner product.

4
 MAT3701/202

QUESTION 2

Let V = M2 2 (C) denote the inner product space over C 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 i i 1 1 1
W = span (S) ; where S = ; :
2 1 1 2 i i

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

z1 z2
(b) Find the formula for T ; expressed as a single matrix in terms of z1 ; z2 ; z3 and
z3 z4
z4 :
1 1
(c) Use T to express A = as A = B + C where B 2 W and C 2 W ? :
1 1

(d) Verify that kT (A)k kAk for all A 2 V:

SOLUTION

1 i i 1 1 1
Let v1 = 2
and v2 = 2
:
1 1 i i

(a)

1 i i i i
hv1 ; v1 i = ;
4 1 1 1 1
1
= (i { + i { + 1 1 + 1 1)
4
1
= ( i2 i2 + 1 + 1)
4
= 1:

Similarly, hv2 ; v2 i = 1:

1 i i 1 1
hv1 ; v2 i = ;
4 1 1 i i
1
= i 1 + i ( 1) + 1 { + 1 ( i)
4
1
= (i i i + i) = 0
4

Thus, S is an orthonormal subset of V .

5
 (b)

z1 z2 z1 z2 z1 z2
T = ; v1 v1 + ; v2 v2
z3 z4 z3 z4 z3 z4
1 i i 1 1 1
= ( iz1 iz2 + z3 + z4 ) + (z1 z2 iz3 + iz4 )
4 1 1 4 i i
1 z1 + iz4 z2 + iz3
=
2 iz2 + z3 iz1 + z4

(c)

A = T (A) + (I T )(A); where T (A) 2 W and (I T )(A) 2 W ?

1 1+i 1+i 1 1 i 1 i
= +
2 1 i 1 i 2 1+i 1+i
= B + C;

where
1 1+i 1+i 1 1 i 1 i
B= 2W and C = 2 W?
2 1 i 1 i 2 1+i 1+i

(d) For all A 2 V;

A = T (A) + (I T )(A);

so
jjAjj2 = jjT (A)jj2 + jj(I T )(A)jj2 jjT (A)jj2 ;

since T (A) and (I T )(A) are orthogonal:

hT (A); (I T )(A)i = hA; T (I T )(A)i

= hA; T (I T )(A)i since T = T (T is an orthogonal projection)
= hA; Oi ; since T 2 = T
= 0:

Alternative solution
This solution is based on the following inequality

(a + b)2 2(a2 + b2 ) for all a; b 2 R: (*)

To prove this, note that

(a + b)2 2(a2 + b2 ) , a2 + 2ab + b2 2a2 + 2b2

, 0 a2 2ab + b2
, 0 (a b)2 ;

and the last inequality is obviously true.

6
 MAT3701/202

Now, from (b),

2 2
z1 z2 1 z1 + iz4 z2 + iz3
T =
z3 z4 4 iz2 + z3 iz1 + z4
1
= jz1 + iz4 j2 + jz2 + iz3 j2 + j iz2 + z3 j2 + j iz1 + z4 j2
4
1
jz1 j + jz4 j)2 + (jz2 j + jz3 j)2 + (jz2 j + jz3 j)2 + (jz1 j + jz4 j)2
4
(since jz1 + iz4 j jz1 j + jiz4 j = jz1 j + jz4 j; and similarly for the others)
1
= (jz1 j + jz4 j)2 + (jz2 j + jz3 j)2
2
1
[2(jz1 j2 + jz4 j2 ) + 2(jz2 j2 + jz3 j2 )]; from (*)
2
= jz1 j2 + jz2 j2 + jz3 j2 + jz4 j2
= jjAjj2 :

Thus
z1 z2 z1 z2 z1 z2
T for all :
z3 z4 z3 z4 z3 z4

QUESTION 3

Consider the system

x y+z = 1
x+y = 2

of linear equations in R3 :

(a) Show that the system has infinitely many solutions. (2)

(b) Find the minimal solution u 2 R3 to the system. (7)

(c) Verify your answer in (b) by finding a parametric representation.

S = fav + w j a 2 Rg

of the solution set S; where v and w are fixed vectors in R3 ; and showing directly that

kuk kav + wk for all a 2 R:

(7)
[16]

7
SOLUTION

(a) The system is consistent since, for example, x = y = z = 1 is a solution and since there are
only two equations in three unknowns, there must be infinitely many solutions.
1 1 1 1
(b) Let A = and b = : The minimal solution is given by u = A x, where x
1 1 0 2
is a solution of AA x = b. So,

AA x = b 2 3
1 1
1 1 1 4 1
() 1 1 5x =
1 1 0 2
1 0
3 0 1
() x=
0 2 2
1
() x = 3 ;
1

and 2 3 2 3
4
1 1 1 3
u=A x=4 1 1 5 3 =4 2
3
5:
1 1
1 0 3

is the minimal solution.

(c)
1 1 1 1 2 0 1 3 R1 ! R1 + R2
!
1 1 0 2 0 2 1 1 R2 ! R2 R1

Let z = a 2 R. Then y = 21 a + 1
2
and x = 1
2
a + 23 .
Thus, 2 3 2 3 2 3 2 3
1
x 2
a + 23 1
2
3
2
4 y 5=4 1
a + 12 5 = a4 1 5+4 1 5 = av + w
2 2 2
z a 1 0
2 1
3 2 3
3
2 2
with v = 4 1
2
5 and w=4 1
2
5 is a parametric representation of S.
1 0

8
 MAT3701/202

Further,
2 1
3 2
a + 32 2 2
2 4
2
5 1 3 1 1
jjav + wjj = 1
2
a + 21 = a+ + a+ + a2
2 2 2 2
a
3 2 5
= a a+
2 2
3 2 5
= a2 a+
2 3 3
" #
2
3 1 14
= a +
2 3 9
2
3 1 7
= a +
2 3 3
7
for all a 2 R
3
= kuk2

Thus, kav + wk kuk for all a 2 R.

QUESTION 4

(a) (cf. Friedberg: Exercise 6.4.5, p. 375.)

Let V be an inner product space over F; and let T be a normal operator on V:

(i) Show that T cI is normal for every c 2 F:

(ii) If T is self–adjoint, show that T cI is self–adjoint if and only if c is real.

(b) (This exercise illustrates Friedberg: Theorem 6.15, p. 371.)

2 3 2 3 2 3
1+i i 0 0 i 2 i
6 i 1+i 0 0 7 6 i 7 6 7
Let A = 6 7 ; and let w1 = 6 7 and w2 = 6 2 + i 7 :
4 0 0 1+i i 5 4 1 i 5 4 0 5
0 0 i 1+i 1+i 0

(i) Show that A is normal.

(ii) Show that kAw1 k = kA w1 k :
(iii) Show that w1 and w2 are eigenvectors of A and find their corresponding eigenvalues 1
and 2 ; respectively.
(iv) Show that A w1 = 1 w1 :

(v) Show that w1 and w2 are orthogonal.

9
SOLUTION

(a) (i)
(T cI)(T cI) = (T cI)(T cI)
= TT cT cT + ccI
= T T cT cT + ccI; since T T = T T
= (T cI)(T cI)
= (T cI) (T cI);
thus (T cI) is normal.
(ii)
(T cI) = T cI
,T cI = T cI
, T cI = T cI; since T = T
, c = c:

(b) (i)
2 32 3
1+i i 0 0 1 i i 0 0
6 i 1+i 0 0 76 i 1 i 0 0 7
AA = 6
4 0
76 7
0 1+i i 54 0 0 1 i i 5
0 0 i 1+i 0 0 i 1 i
2 3
3 2 0 0
6 2 3 0 0 7
= 6
4 0 0
7
3 2 5
0 0 2 3
and
2 32 3
1 i i 0 0 1+i i 0 0
6 i 1 i 0 0 76 i 1+i 0 0 7
AA = 6
4 0
76
54 0
7
0 1 i i 0 1+i i 5
0 0 i 1 i 0 0 i 1+i
2 3
3 2 0 0
6 2 3 0 0 7
= 6
4 0
7
0 3 2 5
0 0 2 3
Thus, A is normal since AA = A A.
2 3 2 3
2+i 2+i
6 2+i 7 6 7
(ii) Aw1 = 6 7 and A w1 = 6 2 + i 7
4 3+i 5 4 1 3i 5
3 i 1 + 3i
so
jjAw1 jj2 = 2j 2 + ij2 + 2j3 + ij2 = 2j2 + ij2 + 2j1 + 3ij2 = jjA w1 jj2
and therefore jjAw1 jj = jjA w1 jj.

10
 MAT3701/202

(iii) From (ii),

2 3 2 3
i 2+i
6 7 6
i 2+i 7
Aw1 = w1 , 6 7 6 7
4 1 i 5=4 3+i 5
1+i 3 i
, i = 2 + i and (1 i) = 3 + i
, = 1 + 2i

Turning to w2 , 2 3
2 i
6 2+i 7
Aw2 = 6
4
7 = w2
5
0
0
Thus, w1 and w2 are eigenvalues of A with corresponding eigenvalues 1 = 1 + 2i and
= 1.
(iv) From (ii), 2 3 2 3
2+i i
6 2+i 7 6 i 7
A w1 = 6 7
4 1 3i 5 = (1 2i) 6
4
7= 1 w1
1 i 5
1 + 3i 1+i
(v)
hw1 ; w2 i = i(2 i) + i( 2 + i) + (1 i)0 + ( 1 + i)0
= i(2 + i) + i( 2 i)
= 0;

hence w1 and w2 are orthogonal.

QUESTION 5

Eliminate the xy–term in

p
5x2 + 2 3xy + 7y 2 = 1 ... ( )
by a rotation of the axes, that is, find a rotation matrix P 2 M2 2 (R) such that ( ), expressed in
terms of x0 ; y 0 defined by
x x0
=P ;
y y0
contains no cross term. Express x; y in terms of x0 ; y 0 , and state the (counterclockwise) angle of
rotation.

SOLUTION

p
p 5 3 x
5x + 2 3xy + 7y 2 =
2
x y p
3 7 y

11
 p
5p 3
Let A = , then
3 7
p
5
p 3
jA Ij =
3 7
= (5 )(7 ) 3
= 2 12 + 32
= ( 4)( 8);
hence the eigenvalues of A are 1 = 4 and 2 = 8.
p p
1p 3 1 3 p
E4 : A 4I = !
3 3 0 0 R2 3R1
so p
1 3
E4 = span :
2 1
p p
p3 3 3 1 R2 p
E8 : A 8I = !
3 1 0 0 R1 + 3R2
so
1 p1
E8 = span :
2 3
Thus, p
1 3 p
1
P = :
2 1 3
(Note that P is a rotation matrix since it is orthogonal and det(P ) = 1:)
x x0
= P
y y0
p
1 3 1p x0
=
2 1 3 y0
p 0 0
1 3x +
= 0
py 0
2 x + 3y
so p 9
3 0 1 0 >
=
x= x + 2y
2 p
1 0 3 0 >
;
y = 2x + y
2
The conic section ( ) in terms of x0 , y 0 becomes
4x02 + 8y 02 = 1:
Since
cos 11
6
sin 11
6
P = ;
sin 11
6
cos 11
6
11
the angle of rotation is (or = 30 ).
6 6

12
 MAT3701/202

QUESTION 6 (Friedberg: Exercise 22, p. 395.)

Let V be a real inner product space.

(a) Prove that any translation on V is a rigid motion. (5)

(b) Prove that the composition of any two rigid motions on V is a rigid motion on V: (5)
[10]

SOLUTION

(a) The general formula of a translation L : V ! V is L(v) = v0 + v, where v0 is a fixed vector in

V . For all v; w 2 V ,

jjL(v) L(w)jj = kv0 + v (v0 + w)k

= jjv wjj;

thus L is a rigid motion.

(b) Suppose f; g : V ! V are rigid motions and v; w 2 V . Then

jjf g(v) f g(w)jj = jjf (g(v)) f (g(w))jj

= jjg(v) g(w)jj; since f is a rigid motion
= jjv wjj; since g is a rigid motion:

Thus f g is a rigid motion.

QUESTION 7

Find the spectral decomposition of

2 3
1+i i 0
A=4 i 1+i 0 5:
0 0 1 + 2i

[12]

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
2 2
= (1 + 2i t)[(1 + i t) i]
= (1 + 2i t)2 (1 t);

13
So the eigenvalues of A are 1 = 1 and 2 = 1 + 2i (with multiplicity two).
2 3 2 3
i i 0 1 1 0 iR1
E1 : A I3 = 4 i i 0 5 ! 4 0 0 0 5 R 2 + R1
1
0 0 2i 0 0 1 2
iR3
so 8 2 39
< 1 1 =
E1 = span p 4 1 5 :
: 2 ;
0
2 3 2 3
i i 0 1 1 0 iR1
E1+2i : A (1 + 2i) I3 = 4 i i 0 5 ! 4 0 0 0 5 R2 R1
0 0 0 0 0 0
so 8 2 3 2 39
< 1 1 0 =
E1+2i = span p 4 5
1 ; 4 0 5 :
: 2 ;
0 1
Let 2 3
p1 p1 0
2 2
P =4 p1
2
p1
2
0 5;
0 0 1
then the spectral decomposition of A is
2 3 2 3
1 0 0 0 0 0
A = P 4 0 0 0 5 P + (1 + 2i) P 4 0 1 0 5 P
0 0 0 0 0 1
2 1 3 2 1 3
p
h i p 0
2 2 p1 p1 0
= 4 p12 5 p12 p12 0 + (1 + 2i) 4 p12 0 5 2 2
0 0 1
0 0 1
2 1 1 3 2 1 1
3
2 2
0 2 2
0
= 4 1 1 5
0 + (1 + 2i) 4 1 1
0 5:
2 2 2 2
0 0 0 0 0 1

QUESTION 8

2 1
Let A = and suppose k k denotes the Euclidean norm.
1 2

(a) Find kAk ; kA 1 k and cond (A) : (6)

(b) Suppose that we have vectors x and x e such that Ax = b; kbk = 1, and kb xk 0:001: Use
Ae
x A 1 bk (the absolute error) and ke
(a) to determine upper bounds for ke x A 1 bk = kA 1 bk (the
relative error). (6)
[12]

Total: [50]

14
 MAT3701/202

SOLUTION

(a)
2 1 2 1 5 4
A A= =
1 2 1 2 4 5
so
5 t 4
jA A tI2 j =
4 5 t
2 2
= (5 t) 4
= (1 t)(9 t);

hence the eigenvalues of A A are 1 = 1 and 2 = 9.

It follows that
p
jjAjj = 2 = 3;
1
jjA 1 jj = p = 1;
1

and
cond(A) = jjAjj jjA 1 jj = 3:

(b)

jje
x A 1 bjj = jjA 1 (Aex) A 1 bjj
jjA 1 jj jjAe
x bjj
0:001

and
x A 1 bjj
jje jjA 1 (Ae
x) A 1 bjj
=
kA 1 bk kA 1 bk
jjAex bjj
cond(A)
kbk
3(0:001) = 0:003

15
 MAT3701/201/2/2015

Tutorial Letter 201/2/2015

LINEAR ALGEBRA

MAT3701

Semester 2

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: 601502

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, 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.

Represent the vectors in V = R2 as columns, with the ordinary addition of column vectors, and let
scalar multiplication over C be defined by
    
c a b c
(a + ib) = , where a, b, c, d ∈ R.
d −b a d

(a) Show that V is a vector space over C.

(b) Find a basis for V.

Solution

(a) Axioms VS1-VS4 (Friedberg, p.7) for addition in V follow from the ordinary addition of column
vectors in R2 . It remains to check axioms VS5-VS8 for scalar multiplication:
VS5:       
c 1 0 c c
1· = = .
d 0 1 d d

2
 MAT3701/201

VS6:
   
c c
((a1 + ib1 )(a2 + ib2 )) = (a1 a2 − b1 b2 + i(a1 b2 + a2 b1 ))
d d
  
a1 a2 − b 1 b 2 a1 b 2 + a2 b 1 c
=
−(a1 b2 + a2 b1 ) a1 a2 − b1 b2 d
    
a1 b 1 a2 b 2 c
=
−b1 a1 −b2 a2 d
     
a1 b 1 a2 b2 c
=
−b1 a1 −b2 a2 d
  
c
= (a1 + ib1 ) (a2 + ib2 ) .
d

VS7:
         
c1 c a b c1 c
(a + ib) + 2 = + 2
d1 d2 −b a d1 d2
     
a b c1 a b c2
= +
−b a d1 −b a d2
   
c1 c
= (a + ib) + (a + ib) 2 .
d1 d2

VS8:
   
c c
((a1 + ib1 ) + (a2 + ib2 )) = (a1 + a2 + i(b1 + b2 ))
d d
  
a1 + a2 b1 + b2 c
=
−(b1 + b2 ) a1 + a2 d
     
a1 b 1 a2 b 2 c
= +
−b1 a1 −b2 a2 d
     
a1 b 1 c a2 b 2 c
= +
−b1 a1 d −b2 a2 d
   
c c
= (a1 + ib1 ) + (a2 + ib2 ) .
d d

Thus all the axioms are satisfied, and therefore V is a vector space over C.

(b) For all a, b ∈ R,       

1 a −b 1 a
(a − ib) = = ,
0 b a 0 b
thus V is generated by [ 10 ] which is clearly linearly independent, and therefore it forms a basis for V.

3
Question 2

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 defined by T (X) = AX.
i −1

(a) Show that T is a linear operator over C.

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

(c) Find a basis for N (T ).

(d) Determine whether or not M2×2 (C) = R (T ) ⊕ N (T ).

Solution

(a) For X, Y ∈ M2×2 (C) and z ∈ C,

T (X + Y ) = A (X + Y )
= AX + AY
= T (X) + T (Y )

and

T (zX) = A (zX)
= z (AX)
= zT (X).

Thus, T is a linear transformation since it satisfies both conditions for a linear transformation.

(b) Since
        
1 0 0 1 0 0 0 0
β = A11 = , A12 = , A21 = , A22 =
0 0 0 0 1 0 0 1

is a basis for M2×2 (C) ,

R (T ) = span {T (A11 ) , T (A12 ) , T (A21 ) , T (A22 )}

= span {AA11 , AA12 , AA21 , AA22 }
       
1 0 0 1 i 0 0 i
= span , , ,
i 0 0 i −1 0 0 −1
   
1 0 0 1
= span ,
i 0 0 i

4
 MAT3701/201

since        
1 0 i 0 0 1 0 i
i = and i = .
i 0 −1 0 0 i 0 −1

Thus,    
1 0 0 1
α= ,
i 0 0 i

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

 
z1 z2
(c) Let X = , where z1 , z2 , z3 , and z4 are complex numbers.
z3 z4

X ∈ N (T ) ⇔ T (X) = 0
⇔ 
AX = 0     
1 i z1 z2 0 0
⇔ =
i −1 z3 z4 0 0
   
z1 + iz3 z2 + iz4 0 0
⇔ =
iz1 − z3 iz2 − z4 0 0
 
z1 + iz3 = 0 z2 + iz4 = 0
⇔ and
iz1 − z3 = 0 iz2 − z4 = 0
⇔ iz1 − z3 = 0 and iz2 − z4 = 0,

since −i (iz1 − z3 ) = z1 + iz3 and −i (iz2 − z4 ) = z2 + iz4

⇔ z3 = iz1 and z4 = iz2 .

Thus,      
z1 z2 1 0 0 1
X= = z1 + z2 .
iz1 iz2 i 0 0 i

A basis for N (T ) is therefore

   
1 0 0 1
β= , .
i 0 0 i

(d) Since α = β, it follows that α ∪ β cannot be a basis for M2×2 (C), and therefore M2×2 (C) is not
the direct sum of R (T ) and N (T ).

Question 3

Let V be a vector space over F with basis β = {v1 , ..., vn }, and let T : V → V be a linear operator.

(a) Show that T 2 = 0V ⇔ [T ]2β = 0. (5)

5
 (b) Let V = M2×2 (C) considered as a vector space over C, and let T : M2×2 (C) → M2×2 (C) be
the linear operator defined by
 
1 i
T (X) = AX, where A = .
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 diagonal, and write down [T ]β . (20)

[25]

Solution

(a) T 2 = 0V ⇔ [T 2 ]β = [0V ]β ⇔ [T ]2β = 0, since [T 2 ]β = [T ]2β (F4: Theorem 2.11) and [0V ]β = 0.

(b) 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  
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
(expansion along the first row)
1−λ i 1−λ i
= − (1 − λ) (1 + λ) +
i −1 − λ i −1 − λ
(expansion along the second row of both)
1−λ i
= [1 − (1 − λ) (1 + λ)]
i −1 − λ
= [1 − (1 − λ) (1 + λ)]2
= λ4 .

Therefore the only eigenvalue of T is 0 with algebraic multiplicity 4.

   
1 0 i 0 1 0 i 0
 0 1 0 i 
  0 1 0 i 
[T ]β − 0I = 
 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,
hence the algebraic and geometric mulitiplicities of λ = 0 are not equal, and therefore T is not
diagonalizable.

Question 4

For  
1 4
A= ∈ M2×2 (R)
2 3
find an expression for An , where n is an arbitrary positive integer. [20]

Solution

1−λ 4
det (A − λI) = = (λ + 1)(λ − 5)
2 3−λ
so the eigenvalues of A are λ = −1 and 5, each with multiplicity 1, and
 
2
E−1 = span (show)
−1

7
and  
1
E5 = span (show).
1
Let  
2 1
X=
−1 1
then 
−1 0
A=X X −1
0 5
so
 n
−1 0
n
A =X X −1
0 5
 
(−1)n 0
=X X −1
0 5n
 1 −1
 
  n
2 1 (−1) 0  3 3
= 
−1 1 0 5n 1 2
3 3
 n n n+1

1 2 · (−1) + 5 2 · (−1) + 2 · 5n
= .
3 (−1)n+1 + 5n (−1)n + 2 · 5n

Question 5

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 V = R (T ) ⊕ N (T ) ⇔ R (T ) ∩ N (T ) = {0}.

(b) Show that R (T ) ∩ N (T ) = {0} if T is a projection.

(c) Give an example of an operator T : C 2 → C 2 such that R (T ) ∩ N (T ) = {0} but T is not a

projection.

(d) Find a formula for a linear operator T : C 3 → C 3 over C such that T is a projection with

R (T ) = span {(1, 1, 1)}

and
N (T ) = span {(1, 1, 0) , (0, 1, 1)}.

8
 MAT3701/201

Solution

(a) Suppose V = R (T ) ⊕ N (T ), then R (T ) ∩ N (T ) = {0} by the definition of direct sum.

Conversely, suppose R (T ) ∩ N (T ) = {0} . Then

dim(R(T ) + N (T )) = dim(R(T )) + dim(N (T )) = dim(V )

so
V = R(T ) + N (T )
and therefore
V = R (T ) ⊕ N (T )
since it is also given that R (T ) ∩ N (T ) = {0}.
(b)

v ∈ R (T ) ∩ N (T ) ⇒ T (v) = v since v ∈ R (T ) and T (v) = 0 since v ∈ N (T )

⇒ v = 0.

Thus R (T ) ∩ N (T ) = {0}.

(c) Define T : C 2 → C 2 by T (v) = −v for all v ∈ C 2 . Then R (T ) ∩ N (T ) = {0} since N (T ) = {0}

but T is not a projection since T 2 = I 6= T.
(d) From the information given we must have

T (1, 1, 1) = (1, 1, 1)
T (1, 1, 0) = (0, 0, 0)
T (0, 1, 1) = (0, 0, 0)

thus

T (1, 0, 0) = T (1, 1, 1) − T (0, 1, 1) = (1, 1, 1)

T (0, 0, 1) = T (1, 1, 1) − T (1, 1, 0) = (1, 1, 1)
T (0, 1, 0) = T (1, 1, 1) − T (1, 0, 0) − T (0, 0, 1) = (−1, −1, −1)

so
T (x, y, z) = xT (1, 0, 0) + yT (0, 1, 0) + zT (0, 0, 1) = (x − y + z, x − y + z, x − y + z).

9
Question 6

This question was not marked.

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

Prove that if a 1–dimensional subspace W of Rn contains a nonzero vector with all nonnegative entries,
then W contains a unique probability vector.
Solution

Suppose v = (v1 , · · · , vn ) is a nonzero vector in W having only nonnegative entries. Then r =

v1 + · · · + vn > 0, and
1
w = v = (w1 , · · · , wn )
r
is a probability vector in W.
To show that w is unique, suppose u = (u1 , · · · , un ) is another probability vector in W. Then u = tw
for some 0 6= t ∈ R since W is one-dimensional and therefore generated by w. Then

1 = u1 + · · · + un = t · w1 + · · · + t · wn = t(w1 + · · · + wn ) = t,

so t = 1 and therefore u = w.

Question 7

 
1 √ 1 −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. (5)
(ii) Calculate the row and column sums of B. (5)
 
0.4 0.6
(b) Let A = .
0.7 0.3

(i) Find the smallest disc centered at the origin of the complex plane in which the eigenvalues
of A lie. (5)
(ii) Find an eigenvalue of A on the circumference of your disc in (i), and a corresponding
eigenvector. (10)
(iii) Find lim Am . (15)
m→∞

[40]

10
 MAT3701/201

Solution

(a)(i)

C1 = {z ∈ C : |z − 1| ≤ 2}
C2 = {z ∈ C : |z − 2| ≤ 3}
C2 = {z ∈ C : |z + 1| ≤ 5}.

(ii)

ρ(B) = max{ρ1 (B), ρ2 (B), ρ3 (B)} = max{3, 5, 6} = 6

ν(B) = max{ν1 (B), ν2 (B), ν3 (B)} = max{8, 3, 3} = 8.

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

(ii)

0.4 − λ 0.6
det (A − λI) = = (λ − 1)(λ + 0.3)
0.7 0.3 − λ
so the eigenvalues of A are λ = 1 and − 0.3. The required eigenvalue is therefore λ = 1, and
 
1
E1 = span (show).
1

(iii) We use the fact that At is a regular transition matrix, and

lim Am = ( lim (At )m )t .

m→∞ m→∞

To find an eigenvector of At corresponding to the eigenvalue λ = 1 we solve the homogeneous linear

system with coefficient matrix  
−0.6 0.7
,
0.6 −0.7
   
7 1 7
a solution of which is ; hence 13 is a probability vector which is also an eigenvector corre-
6 6
sponding to λ = 1, and therefore  
t m 1 7 7
lim (A ) =
m→∞ 13 6 6
so  
m 1 7 6
lim A = .
m→∞ 13 7 6

11
Question 8

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. (10)

(b) Find the characteristic polynomial of TW . (5)

[15]
Solution

(a)
   
0 0 00 0 0 0 0
 1 0 00 
, T (A) = A2 =  0 0 0 0 

A = 
 0
,
1 10   1 1 1 0 
0 0 01 0 0 0 1
 
0 0 0 0
 0 0 0 0 
T 2 (A) = T (T (A)) = T A2 = A3 =   = A2 = T (A).


 1 1 1 0 
0 0 0 1

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

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

Total [50]

12
 MAT3701/202/2/2015

Tutorial Letter 202/2/2015

LINEAR ALGEBRA

MAT3701

Semester 2

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
Solutions
UNIQUE ASSIGNMENT NUMBER: 601533

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 of
a given year (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 should be fully motivated.

For this assignment, questions 2, 4, 6, and 8 will be marked.

Question 1

This question was not marked.

It is still important to work through the solution and compare it with your own attempt.

Let V be an inner product space. Prove that

(a) kx ± yk2 = kxk2 ± 2R hx, yi + kyk2 for all x, y ∈ V, where R hx, yi denotes the real part of the
complex number hx, yi.

(b) |kxk − kyk| ≤ kx − yk for all x, y ∈ V.

Solution

(a)

kx ± yk2 =hx ± y, x ± yi
=hx, xi ± hx, yi ± hy, xi + hy, yi
=hx, xi ± hx, yi ± hx, yi + hy, yi
=hx, xi ± 2Rhx, yi + hy, yi
= kxk2 ± 2R hx, yi + kyk2 .

2
 MAT3701/202

(b)
kxk = k(x − y) + yk ≤ kx − yk + kyk
so
kxk − kyk ≤ kx − yk . (1)
Similarly
kyk = kx − (x − y)k ≤ kxk + kx − yk
so
kyk − kxk ≤ kx − yk . (2)
Combining (1) and (2) we have
±(kxk − kyk) ≤ kx − yk
and therefore
|kxk − kyk| ≤ kx − yk .

Question 2

Prove that if {w1 , w2 , ..., wn } is an orthogonal set of nonzero vectors, then the vectors v1 , v2 , ..., vn
derived from the Gram–Schmidt process satisfy vi = wi for i = 1, 2, ..., n. Hint: Use mathematical
induction. [10]

Solution

Clearly v1 = w1 . Assume the result is true for k < n, that is

v1 = w1 , · · · , vk = wk .
Then
k
X hwk+1 , vj i
vk+1 = wk+1 − vj
j=1
kvj k2
k
X hwk+1 , wj i
= wk+1 − vj
j=1
kwj k2
= wk+1 , since hwk+1 , wj i = 0 for 1 ≤ j ≤ k.
Thus the result follows by mathematical induction.

Question 3

This question was not marked.

It is still important to work through the solution and compare it with your own attempt.

Consider the system

x−y+z = 1
x−y+z = 2

3
of linear equations in R3 , which we write as
 
x  
1
A y =
 
2
z
where  
1 −1 1
A= .
1 −1 1

(a) Show that the system has no solution.

(b) Find a least squares approximate solution u ∈ R3 to the system (see SG: Section 12.6).
(c) Verify your answer in (b) by showing directly that
     
x−y+z 1 1
− ≥ Au − for all x, y, z ∈ R.
x−y+z 2 2

Solution

(a) Subtracting the first equation from the second yields

0x − 0y + 0z = 1,

thus the system has no solutions.

(b) Use the method described in Section 12.6 of the Study Guide:
   
2 −2 2   3
∗ ∗ 1
A A = −2 2 −2 and A
  = −3 ,

2
2 −2 2 3
 
1
and it follows that a least squares approximate solution u is given by A∗ Au = A∗ , for example
2
3
2
u =  0  (show).
0

(c) This solution is based on the inequality

1
a2 + b2 ≥ (a − b)2 for all a, b ∈ R, (*)
2
and which can be established as follows:
1
a2 + b2 ≥ (a − b)2 ⇔ 2a2 + 2b2 ≥ a2 − 2ab + b2
2
⇔ a2 + 2ab + b2 ≥ 0
⇔ (a + b)2 ≥ 0,

4
 MAT3701/202

and the last inequality is obviously true.

Now
    2
x−y+z 1
− = (x − y + z − 1)2 + (x − y + z − 2)2
x−y+z 2
1
≥ [(x − y + z − 1) − (x − y + z − 2)]2 from (∗)
2
1
≥
2
  2
 3 
2 1
=  −
3 2
2
  2
1
= Au − ,
2
thus the result follows.

Question 4

Let  
z1 z2
A= ∈ M2×2 (C) .
z2 z3

(a) Show that A is normal iff z1 z̄2 + z2 z̄3 is real. (5)

(b) Show from (a) that A is normal if z1 = z3 . (3)
(c) Suppose z1 = z3 and z2 6= 0.
(i) Find the eigenvalues λ1 and λ2 of A with corresponding eigenvectors v1 and v2 , respectively;(5)
(ii) Show that A∗ v1 = λ̄1 v1 ; (3)
(iii) Show that v1 and v2 are orthogonal; (1)
∗
(iv) Show that kAv1 k = kA v1 k. (3)
[20]

Solution

(a)     
z z z̄1 z̄2 z1 z̄1 + z2 z̄2 z1 z̄2 + z2 z̄3
AA = 1 2
∗
=
z2 z3 z̄2 z̄3 z2 z̄1 + z3 z̄2 z2 z̄2 + z3 z̄3
and     
∗ z̄1 z̄2 z1 z2 z̄1 z1 + z̄2 z2 z̄1 z2 + z̄2 z3
A A= = .
z̄2 z̄3 z2 z3 z̄2 z1 + z̄3 z2 z̄2 z2 + z̄3 z3

AA∗ = A∗ A ⇔ z1 z̄2 + z2 z̄3 = z̄1 z2 + z̄2 z3

⇔ z1 z̄2 + z2 z̄3 = z1 z̄2 + z2 z̄3
⇔ z1 z̄2 + z2 z̄3 is real.

5
(b) If z1 = z3 then
z1 z̄2 + z2 z̄3 = z1 z̄2 + z2 z̄1 = z1 z̄2 + z1 z̄2 ,
which is real, so A is normal according to (a).

(c)(i)
z1 − λ z2
det(A − λI) = = (z1 − λ)2 − z22 = (λ − z1 − z2 )(λ − z1 + z2 ),
z2 z1 − λ
therefore λ1 = z1 + z2 and λ2 = z1 − z2 .

  
−z2 z2 1
A − λ1 I = , so v1 =
z2 −z2 1

and 
  
z2 z2 1
A − λ2 I = , so v2 = .
z2 z2 −1

(c)(ii)       
∗ z̄1 z̄2 1 z̄1 + z̄2 1
A v1 = = = (z1 + z2 ) ,
z̄2 z̄1 1 z̄1 + z̄2 1
as required.

(c)(iii)    
1 1
hv1 , v2 i = h , i = 0,
1 −1
thus v1 and v2 are orthogonal.

(c)(iv)
  2
z1 + z2
kAv1 k2 =   = 2(z1 + z2 )(z1 + z2 ) = 2 |z1 + z2 |2
z1 + z2
and   2
z1 + z2
kA∗ v1 k2 =   = 2(z1 + z2 )(z1 + z2 ) = 2 |z1 + z2 |2 ,
z1 + z2
thus kAv1 k = kA∗ v1 k .

Question 5

This question was not marked.

It is still important to work through the solution and compare it with your own attempt.

Prove that if T is a unitary operator on a finite–dimensional inner product space V, then T has a
unitary square root; that is, there exists a unitary operator U such that T = U 2 .

Solution

6
 MAT3701/202

By F4:Theorem 6.18 and its corollaries, V has an orthonormal basis

B = {v1 , · · · , vn }

of eigenvectors of T such that the corresponding eigenvalues are all of absolute value 1; say T√(vi ) = λi vi
with |λi | = 1, 1 ≤ i ≤ n. Let U : V → V be the unique linear operator such that U (vi ) = λi vi , 1 ≤
i ≤ n. Then T = U 2 and it follows from the same result above that U is unitary.

Question 6

Let T and U be orthogonal operators on R2 . Use F4:Theorem 6.23 to prove the following results.

(a) If T and U are both reflections about lines through the origin, then U T is a rotation. (5)

(b) If T is a rotation and U is a reflection about a line through the origin, then both U T and T U
are reflections about lines through the origin. (5)

[10]

Solution

Note that the product of two orthogonal projections is orthogonal.

(a) Since det(T ) = det(U ) = −1 and det(U T ) = det(U ) det(T ) = 1, it follows from F4:Theorem 6.23
that U T is a rotation.
(b) Since det(T ) = 1, det(U ) = −1, and det(U T ) = det(T U ) = det(U ) det(T ) = −1, it follows from
F4:Theorem 6.23 that both U T and T U are reflections.

Question 7

This question was not marked.

It is still important to work through the solution and compare it with your own attempt.

It is given that A ∈ M3×3 (C) is a normal matrix with eigenvalues 1 and i and corresponding
eigenspaces
√  1 √ 
  
1
E1 = span 1, 1, 2 , 1, 1, − 2
2 2
and  
1
Ei = span √ (−1, 1, 0) .
2

(a) Find the spectral decomposition of A.

(b) Find A.

7
Solution

(a)  1 1  1 1
2 2 0 2 2
 12 1 √1
 
 2
 
2
1 1
  1 1 
P1 =   2 2 0
2 2
 = 
−1
  1 1
  √  
2 2 2
√1 −1
√
2 2
0 0 1
and  −1  1 −1
 
√ 0
2 2 2
 h i  
 −1
√1
1  −1 1

Pi = 
 √2
 √
 2 2
0 =
2 2
0

   
0 0 0 0
The spectral decomposition is

A = 1 · P1 + i · Pi
−1
1 1  1 
2 2
0 2 2
0
   
 + i ·  −1
1 1   1

=1· 2 2  0 2 2
0

   
0 0 1 0 0 0

(b) 1 i 1 i

2
+ 2 2
− 2
0
 
1 i 1 i

2 −
A= 2 2
+ 2
0.
 
0 0 1

Question 8
 
3 −1
Let A = and suppose k·k denotes the Euclidean norm.
1 −3

(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). (3)

[10]
Solution

(a) 
   
∗ 3 1 3 −1 10 −6
A A= = ,
−1 −3 1 −3 −6 10

8
 MAT3701/202

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
−1
= ≤ cond (A) ≤ 2(0.001) = 0.002.
kA bk kxk kbk
Total [50]