The Examination of Semester 1, 2016-17

0 1
1 0 1 1
B C
B0 k 1 k
Let A = B
C
1. [14 marks] C where k is a constant.
@k k 2 0 A
k 0 k 0
(a) Use Gauss-Jordan Elimination to reduce A to the reduced row-echelon form.
(Write down the elementary row operations clearly.)
(b) Find a basis for the nullspace space of A.

2. [12 marks] Let V = f (a + b 2c; 2b c; 3c + d; a + 3b + d) j a; b; c; d 2 R g.

(a) Show that V is a subspace of R4 .
(b) Find a basis for V and determine the dimension of V .
(c) Let W = f (1 + a + b 2c; 2b c; 3c + d; 1 + a + 3b + d) j a; b; c; d 2 R g.
(i) Is the zero vector contained in W ? Justify your answer.
(ii) Is W a subspace of R4 ?

0 1
4 0 2 2
B
B0 4 2 2C
Let B = B
C
3. [18 marks] C.
@0 0 2 2A
0 0 2 2

(a) Find an invertible matrix P and a diagonal matrix D so that P 1 BP = D.

(b) Write down a matrix C such that C 2 = B .
(You can express your answer in the form PXP 1 where X is a 4  4 matrix
and P is the invertible matrix obtained in (a).)

4. [22 marks] (All vectors in this question are written as column vectors.)
0 1 0 1 0 1 0 1 0 1 0 1
1 0 0 1 0 0
Let u1 = 1 , u2 = 1 , u3 = 1 , and e1 = 0 , e2 = 1 , e3 = 0A.
@ A @ A @ A @ A @ A @
1 1 1 0 0 1
(a) (i) Find the reduced row-echelon form of the following 3  6 matrix

u1 u2 u3 e1 e2 e3
:
(ii) Let S = fu1 ; u2 ; u3 g. You can assume that S is a basis for R3 .
Write down the coordinate vectors (e1 )S , (e2 )S , (e3 )S .

1
 (b) De ne a linear transformation T : R3 ! R3 such that

T (x) = c1 u1 + c2 u2 for x = c1 u1 + c2 u2 + c3 u3 2 R 3 :
(i) Find the standard matrix for T .
(ii) Determine the rank and nullity of T .
(iii) Explain why T (x) is the orthogonal projection of x onto V = spanfu1 ; u2 g.

5. [17 marks] (All vectors in this question are written as column vectors.)
Let A be a square matrix of order n.
(a) Let E = fe1 ; e2 ; : : : ; en g be the standard basis for Rn .
Show that Aej = the j th column of A.
(b) Suppose Am = 0 and Am 1 6= 0 for some integer m  2.
(i) Show that there exists at least one vector u 2 Rn such that Am 1 u 6= 0.
(ii) Show that fu; Au; : : : ; Am 1 ug is linearly independent where u is the
vector obtained in Part (i).
(c) Prove that if An+1 = 0, then An = 0.

6. [17 marks] (All vectors in this question are written as column vectors.)
(a) Let B be a 2  2 symmetric matrix and let u, v be two eigenvectors of B
associated with the eigenvalues  and  respectively.
(i) Show that if  6= , then u
v = 0.
! !
p1 p1
(ii) Suppose u = 2 ,v= 2 ,  = 1 and  = 3.
p1 p1
2 2
Find B .
(b) Let C be a symmetric matrix of order n with a characteristic polynomial

( 1 )( 2 )


( n )

where 1  2 


 n .
x 2 Rn, 1  xxTCx
T
Prove that for any nonzero vector
x  n.

2
Solution:

0 1 0 1 0 1
1 0 1 1
B C R3 kR1 B10 0 1 1
C R3 R2 B01 0 1 1
C
B0 k 1 k k 1 k k 1 k
1. (a) B
@k k 2 0
C
C ! B B
k 2 k k
C
C ! B B
0 1 k
C
C

k 0 k 0
A
R4 kR1 @00 0 0 k
A @0
0 0 0
0
k
A

Case 1: k = 0.
0 1 0 1
1 0 1 1
B
0C
R3 R2 B10 0 0 1
0C
B0 0 1 0 1
B
@0 0 1 0A
C
C ! B B
C
C

0 0 0 0
R1 R2 @00 0
0
0
0
0A
0

Case 2: k = 1.
0 1 0 1 0 1
1 0 1 1
B
1C
R3 $ R4 B10 0 1 1
1C
R1 + R3 B10 0 1 0
0C
B0 1 1 1 1 1 1
B
@0 0 0 0A
C
C ! B B
@0 0 0 1A
C
C ! B B
C
C

0 0 0 1 0 0 0 0
R2 R3 @00 0
0
0
0
1A
0

Case 3: k 6= 0 and k 6= 1.
0 1 1 R2 01 0 1 0 1
1 0 1 1 1 1
B
kC
k
B 1 1C
R1 + R4 B10 0 1 0
1 C
B0 k 1 C 1 1 k R3 B0 1 1 k 0C
B
@0 0 1 k
C
0 A ! @0 0
B k
1 0A
C
C ! B B C
1 R2 R4 @00 0 1 0A
k R4
0 0 0 k 0 0 0 1 0 0 1
0 1
R1 R3 B10 0
1
0
0
0
0C
! B B
C
C
R2 k R3 @00
1 0
0
1
0
0A
1
Alternatively , for this case, since A is invertible, the reduced row-
echelon form is an identity matrix.

(b) Case 1: k = 0.
The general solution for Ax = 0 is
0 1 0 1 0 1
t 0 1
B C B C B C
sC B1C B0C
x=B
B C = sB C + tB C
@0A @0A @0A
for s; t 2 R:
t 0 1

Thus f (0; 1; 0; 0)T ; (1; 0; 0; 1)T g is a basis for the nullspace of A.

3
 Case 2: k = 1.
The general solution for Ax = 0 is
0 1 0 1
t 1
B C B
B tC 1C
x=B B
C = tB
@ t A @ 1A
C
C for s; t 2 R:
0 0

Thus f ( 1; 1; 1; 0)T g is a basis for the nullspace of A.

Case 3: k 6= 0 and k 6= 1.
The linear system Ax = 0 has only the trivial solution, i.e. the solution
space is the zero space.
The basis for the nullspace of A is the empty set, ;.

2. (a) Since
(a + b 2c; 2b c; 3c + d; a + 3b + d)
= a(1; 0; 0; 1) + b(1; 2; 0; 3) + c( 2; 1; 3; 0) + d(0; 0; 1; 1);

V = spanf (1; 0; 0; 1); (1; 2; 0; 3); ( 2; 1; 3; 0); (0; 0; 1; 1) g and hence is a

subspace of R4 .

(b) Method 1:
0 1 0 1
1 0 0 1 1 0 0 1
B Guassian
1 2 0 3C B
B0 2 0 2C
B
B
@ 2 1 3
C
C
0A
! B
@0 0 3 3A
C
C
Elimination
0 0 1 1 0 0 0 0

f (1; 0; 0; 1); (0; 2; 0; 2); (0; 0; 3; 3) g is a basis for V and dim(V ) = 3.

Method 2:
0 1 0 1
1 1 2 0 1 1 2 0
B Guassian
B0 2 1 0C B
B0 2 1 0C
B
@0 0 3 1A
C
C ! B
@0 0 3 1A
C
C
Elimination
1 3 0 1 0 0 0 0

f (1; 0; 0; 1); (1; 2; 0; 3); ( 2; 1; 3; 0) g is a basis for V and dim(V ) = 3.

(c) (i) Yes.

(1 + a + b 2c; 2b c; 3c + d; 1 + a + 3b + d) = (0; 0; 0; 0)

4
 8
>
>
>
a + b 2c = 1
<
2b c = 0
, > 3c + d = 0
>
>
:
a + 3b +d = 1:
0 1 0 1
1 1 2 0 1 1 1 2 0 1
B Guassian
B0 2 1 0 0C B
B0 2 1 0 0C
B
@0 0 3 1 0A
C
C ! B
@0 0 3 1 0A
C
C
Elimination
1 3 0 1 1 0 0 0 0 0
The linear system is consistent, i.e. there exists real numbers a, b, c, d
such that (1 + a + b 2c; 2b c; 3c + d; 1 + a + 3b + d) = (0; 0; 0; 0).
So the zero vector is contained in V .
(ii) Yes.
Remark: Actually, W = V .

 4 0 2 2
0  4 2 2
3. (a) det(I B) = 0 0  2 2
0 0 2  2
 2 2
= ( 4)2
2  2
=  ( 4)3 .
The eigenvalues of B are 0 and 4.
0 1 0 1
4 0 2 2 0
B C B C
0 4 2 2C B0C
 For  = 0, (I B)x = , C x = B C.
B
0 B
@ 0 0 2 2A @0A
0 0 2 2 0
0 1 0 1
4 0 2 2 0 1 0 0 1 0
B C Guass-Jordan B C
0 4 2 2 0 0 1 0 1 0
B
B
@ 0 0 2 2 0
C
C
A
B
B
@
! 0 0 1 1 0
C
C
A
Elimination
0 0 2 2 0 0 0 0 0 0
The general solution is
0 1 0 1
t 1
B
tC B
1C
x B
=B
@
C B
C = tB
tA @
C for t 2 R:
C
1A
t 1
So f (1; 1; 1; 1)T g is a basis for E0 .

5
 0 1 0 1
0 0 2 2 0
B C B C
B0 0 2 2C B0C
 For  = 4, (I B)x = 0 , B C x = B C.
@0 0 2 2A @0A
0 0 2 2 0
0 1 0 1
0 0 2 2 0 0 0 1 1 0
B C Guass-Jordan B C
0 0 2 2 0 0 0 0 0 0
B
B
@ 0 0 2 2 0
C
C
A
!B
B
@ 0 0 0 0 0
C
C
A
Elimination
0 0 2 2 0 0 0 0 0 0
The general solution is
0 1 0 1 0 1 0 1
r 1 0 0
B C B C B C B C
sC B0C B1C B0C
x=B
B C = rB C + sB C + tB C for r; s; t 2 R:
@tA @0A @0A @1A
t 0 0 1
So f (1; 0; 0; 0)T ; (0; 1; 0; 0)T ; (0; 0; 1; 1)T g is a basis for E4 .
0 1 0 1
1 1 0 0 0 0 0 0
B C B
1 0 1 0C B0 4 0 0C
Let P C and D = B P 1BP = D.
B C
=B C. Then
@ 1 0 0 1A @0 0 4 0A
1 0 0 1 0 0 0 4
0 1
0 0 0 0
B
B0 2 0 0C
(b) For example, let C = PB CP . Then
1 C 2 = B.
C
@0 0 2 0A
0 0 0 2

0 1 0 1
1 0 0 1 0 0 Guass-Jordan 1 0 0 1 0 0
4. (a) (i) @1 1 1 0 1 0A ! B
@0 1 0 1 1
2
1 C
2 A
1 1 1 0 0 1 Elimination 0 0 1 0 1 1
2 2
(ii) (e1 )S = (1; 1; 0), (e2 )S = (0; 12 ; 12 ), (e3 )S = (0; 21 ; 1 ).
2
0 1 0 1 0 1
1 0 1
(b) (i) T (e1 ) = u1 u2 = @1A @1A = @0A.
1 1 0
0 1 0 1
0 0
T (e2 ) = 0u1 + 2 u2 = 2 @1A = @ 21 C
1 1 B
A.
1 1
2
0 1 0 1
0 0
T (e3 ) = 0u1 + 2 u2 = 2 1 = @ 2 A.
1 1 @ A B1C

1 1
2
6
 Thus the standard matrix for T is 0 1
1 0 0

B
T (e1 ) T (e2 ) T (e3 ) = @0 1
2
1C :
2A
0 1 1
2 2
(ii) From the de nition of T , R(T ) = spanfu1 ; u2 g. It is obvious that
fu1; u2g is linearly independent and hence is a basis for R(T ).
So rank(T ) = dim(R(T )) = 2
and nullity(T ) = dim(R3 ) rank(T ) = 3 2 = 1.
(iii) Since u3
u1 = 0 and u3
u2 = 0, u3 is orthogonal to V .
For any x = c1 u1 + c2 u2 + c3 u3 2 R3 ,

x = c1u1 + c2u2 + c3u3 = T (x) + c3u3:

As T (x) = c1 u1 + c2 u2 2 V and c3 u3 is orthogonal to V , T (x) is the
orthogonal projection of x onto V .

5. (a) Let A = a1 a2


a n where a j is the j th column of A.
0 1
0
B .. C
B.C
B C
B0C

B C
Ae j = a1 a2


a B C
n B1C the j th coordinate
B0C
B C
B.C
@ .. A
0
= 0a1 +


0aj 1 + 1aj + 0aj +1 +


+ 0an = aj .

(b) (i) Since Am 1 6= 0, Am 1 has at least one nonzero column, say, the j th
column is nonzero.
Let u = e . Then by (a), Am 1 e = the j th column of Am 1 6= 0.
j j

(ii) Observe that Am+k = Am Ak = 0Ak = 0 for all integer k  0.

Consider the vector equation

c1 u + c2 Au +


+ cm Am 1 u = 0: ()

Pre-multiplying Am 1 to both sides of (), we have

Am 1(c1u + c2Au +


+ cnAm 1u) = Am 10

) c1Am 1u + c2Amu +


+ cnA2m 2u = 0

) c1 Am 1 u = : 0

7
 Since Am 1u is a nonzero vector, c1 = 0.
Suppose we have already shown that c1 = 0, c2 = 0, : : : , ci = 0 for some
i with 1  i < m.
Pre-multiplying Am i 1 to both sides of (), we have

Am i 1 (c Ai u + c Ai+1 u +


+ c Am 1 u) = Am
i+1 i+2 n
i 10
) ci+1 A u + ci+2 A u +


+ cn A i 2 u = 0
m 1 m 2m

) ci+1 Am 1 u = 0:

Since Am 1u is a nonzero vector, ci+1 = 0.

By mathematical induction, we have shown that c1 = 0, c2 = 0, : : : ,
cm = 0, i.e. () has only the trivial solution.
So fu; Au; : : : ; Am 1 ug is linearly independent.

(c) (Prove by contradiction.)

Assume that An 6= .
0

Applying (b) with m = n+1, there exists u 2 Rn such that fu; Au; : : : ; Anug
is linearly independent.
But this contradicts the fact that a set of n + 1 vectors in Rn cannot be
linearly independent (by Theorem 3.2.7).
So we conclude that our assumption is wrong and hence An = 0.

6. (a) (i) Since BT = B,

u
(Bv) = uTBv = uTBTv = (Bu)Tv = (Bu)
v:
By Bu = u and Bv = v ,

(u
v ) = (u)
v = (Bu)
v = u
(Bv ) = u
(v ) = (u
v )

) ( )(u
v) = 0:
As  6= , u
v = 0.
! !
p1 p1 1 0 p1 p1 T 2 1
(ii) B= 2
p1 p1
2 2
p1 p1
2 = .
2 2 0 3 2 2 1 2

(b) Method 1:

Since C is a symmetric matrix, there exists an orthonormal basis fu1 ; u2 ; : : : ; ug

n

for Rn such that Cui = i ui for i = 1; 2; : : : ; n.

For any nonzero x 2 Rn , we can write x = c1 u1 + c2 u2 +


+ cn un for some
c1 ; c2 ; : : : ; cn 2 Rn .

8
Since fu1 ; u2 ; : : : ; u g is orthonormal, u
u
n i i = 1 and ui
uj = 0 for i 6= j .
So
xTx = x
x
= (c1 u1 + c2 u2 +


+ cn u )
(c1 u1 + c2 u2 +


+ cn u
n n )
= c21 + c22 +


+ c2n :
On the other hand,
xTCx = x
(Cx)
= (c1 u1 + c2 u2 +


+ cn u )
(c1 Cu1 + c2 Cu2 +


+ cn Cu )
n n

= (c1 u1 + c2 u2 +


+ cn u )
(1 c1 u1 + 2 c2 u2 +


+ n cn u )
n n

= 1 c21 + 2 c22 +


+ n c2n :

By 1  2 


 n ,
1 (c21 + c22 +


+ c2n )  1 c21 + 2 c22 +


+ n c2n  n (c21 + c22 +


+ c2n )

) 1  1c1c2++c22c2++





++c2ncn  n
2 2 2
1 2 n

) 1 
xTCx  n:
xT x
Method 2:

Since C is a symmetric matrix, there exists an orthogonal matrix P such

that 0 1
1 0
B 2 C
P TCP = B
B
@ ...
C
C:
A
0 n
0 1
y1
By C
B 2C
De ne y = P T x, i.e. x = Py. Write y = B .. C. Then
@.A
yn
xTCx = yTP TCPy = yTP TCPy = 1y12 + 2y22 +


+ nyn2 :
xTx yTP TPy yTy y12 + y22 +


+ yn2
By 1  2 


 n ,
1 (y12 + y22 +


+ yn2 )  1 y12 + 2 y22 +


+ n yn2  n (y12 + y22 +


+ yn2 )

) 1  1y1y2++y2y2 2++





++y2 nyn  n

2 2 2
1 2 n

) 1 
xTCx  n:
xT x