NATIONAL UNIVERSITY OF SINGAPORE

Department of Mathematics

MA1522 Linear Algebra for Computing Tutorial 6

     
1 0 0
1. (a) Let u1 =  2 , u2 = 2, u3 = −1. Show that S = {u1 , u2 , u3 } forms a
−1 1 3
3
basis for R .
Solution: We have
1 0 0
2 2 −1 = 7 ̸= 0.
−1 1 3
Thus S = {u1 , u2 , u3 } is a basis for R3 .
 
1
(b) Suppose w = 1. Find the coordinate vector of w relative to S.

1
Solution: We have
     
1 0 0 1 1 0 0 1 1
RREF
 2 2 −1 1  −−−−→  0 1 0 −1/7  ⇒ [w]S = −1/7 .
−1 1 3 1 0 0 1 5/7 5/7
   
1 −1
3
(c) Let T = {v1 , v2 , v3 } be another basis for R where v1 = 5 , v2 =
   3 ,
  4 7
2
v3 = 2. Find the transition matrix from T to S.

4
Solution:
   
1 0 0 1 −1 2 1 0 0 1 −1 2
RREF
 2 2 −1 5 3 2  − −−−→  0 1 0 2 3 0 
−1 1 3 4 7 4 0 0 1 1 1 2
 
1 −1 2
So, the transition matrix P from T to S is P = 2
 3 0 .
1 1 2

(d) Find the transition matrix from S to T .

Solution:  
3/4 1/2 −3/4
Q = P−1 = −1/2 0 1/2  .
−1/8 −1/4 5/8
 (e) Use the vector w in Part (b). Find the coordinate vector of w relative to T .
Solution:
    
3/4 1/2 −3/4 1 1/7
[w]T = Q[w]S = −1/2 0 1/2  −1/7 = −1/7 .
−1/8 −1/4 5/8 5/7 5/14

2. Let V be a subspace of Rn and S = {u1 , u2 , u3 } be a basis for a subspace V . Define

T = {v1 , v2 , v3 } where
v1 = u1 + u2 + u3 , v2 = u2 + u3 and v3 = u2 − u3 .
(a) Show that T is a basis for V .
Solution: By construction, vi ∈ V for i = 1, 2, 3, and hence, span(T ) ⊆ V . Next,
|T | = 3 = dim(V ). Finally, suppose

0 = c1 v 1 + c2 v 2 + c3 v 3
= c1 (u1 + u2 + u3 ) + c2 (u2 + u3 ) + c3 (u2 − u3 )
= c1 u1 + (c1 + c2 + c3 )u2 + (c1 + c2 − c3 )u3

Then since S is linearly independent,


 c1 = 0
c1 + c2 + c3 = 0
c1 + c2 − c3 = 0


which has only the trivial solution c1 = c2 = c3 = 0. Hence, T is linearly indepen-

dent. Thus, T is a basis.

(b) Find the transition matrix from S to T .

Solution: Observe that by construction,
     
1 0 0
[v1 ]S = 1 , [v2 ]S = 1 , [v3 ]S =  1  .
1 1 −1
Hence, the transition matrix P from T to S is
 

1 0 0
P = [v1 ]S [v2 ]S [v3 ]S = 1 1 1  .
1 1 −1
Hence, the transition matrix Q from S to T is
 
1 0 0
Q = P−1 = −1 1/2 1/2  .
0 1/2 −1/2
    
1 −1 1 2
3. (a) Let A =  1 1 −1 and b = 1. Is b in the column space of A?
 
−1 −1 1 0
If it is, express it as a linear combination of the columns of A.

Solution:    
1 −1 1 2 1 0 0 0
 1 1 −1 1  −→  0 1 −1 0  .
−1 −1 1 0 0 0 0 1
Thus b is not a linear combination of the columns of A.
 
1 9 1

(b) Let A = −1 3 1 and b = 5, 1, −1 . Is b in the row space of A? If it is,
1 1 1
express it as a linear combination of the rows of A.

Solution: Note that b is in the row space of A if and only if bT is in the column
space of AT . Hence we are solving for
   
1 −1 1 5 1 0 0 1
 9 3 1 1  −→  0 1 0 −3  .
1 1 1 −1 0 0 1 1

We get b = (5, 1, −1) = (1, 9, 1) − 3(−1, 3, 1) + (1, 1, 1).

 
1 2 0 1
0 1 2 1 4
(c) Let A = 
1 2 1 3. Is the row space and column space of A the whole R ?


0 1 2 2
Solution:    
1 2 0 1 1 0 0 0
0 1 2 1
  0 1 0 0

1 −→  .
2 1 3 0 0 1 0
0 1 2 2 0 0 0 1
Thus the column space of A is the whole R4 . Since A is invertible if and only if
AT is, the row space must also be the whole R4 .

4. For each of the following matrices A,

(i) Find a basis for the row space of A.

(ii) Find a basis for the column space of A.
(iii) Find a basis for the nullspace of A.
(iv) Hence determine rank(A), nullity(A) and verify the dimension theorem for matri-
ces.
(v) Is A full rank?
  
1 2 5 3
 1 −4 −1 −9
 
−1 0 −3 1 
(a) A =  
2 1 7 0
0 1 1 2
Solution:  
1 0 3 −1
0 1 1 2
 
A −→ 
0 0 0 0.
0 0 0 0
0 0 0 0

(i) A basis for the row space is {(1, 0, 3, −1), (0, 1, 1, 2)}.

(ii) A basis for the column space is {(1, 1, −1, 2, 0)T , (2, −4, 0, 1, 1, )T }.

(iii) A basis for the nullspace is {(−3, −1, 1, 0)T , (1, −2, 0, 1)T }.

(iv) rank(A) = 2, nullity(A) = 2. Since rank(A) + nullity(A) = 2 + 2 = 4 which

is the number of columns of A, we have verified the dimension theorem for
matrices.

(v) rank(A) = 2 < min{4, 5}. A is not full rank.

 
1 3 7
2 1 8
 
(b) A =  3 −5 −1 .

2 −2 2 
1 1 5
Solution:  
1 0 0
0 1 0
 
A −→ 
0 0 1
.
0 0 0
0 0 0

(i) A basis for the row space is {(1, 0, 0), (0, 1, 0), (0, 0, 1)}.

(ii) A basis for the column space is {(1, 2, 3, 2, 1)T , (3, 1, −5, −2, 1)T , (7, 8, −1, 2, 5)T }.

(iii) The basis for the nullspace is the empty set.

(iv) rank(A) = 3, nullity(A) = 0. Since rank(A) + nullity(A) = 3 + 0 = 3 which

is the number of columns of A, we have verified the dimension theorem for
matrices.

(v) rank(A) = 3 = min{3, 5}. A is full rank.

5. Let W be a subspace of R5 spanned by the following vectors
       
1 2 0 2
−2 −5 5 1
       
 0  , u2 −3 , u3 = 15 , u4 = 15 .
u1 =        
0 −2 10 8
3 6 0 6

(a) Find a basis for W .

Solution:    
1 −2 0 0 3 1 0 6 0 3
2 −5 −3 −2 6 RREF 0 1 3 0 0
  −−−−→  
0 5 15 10 0 0 0 0 1 0
2 1 15 8 6 0 0 0 0 0
      

 1 0 0 
0 1 0

      


6 , 3 , 0 forms a basis for W .
     


 0 0 1 
 
3 0 0
 

(b) What is dim(W )?

Solution: From (a), dim(W ) = 3

(c) Extend the basis W found in (a) to a basis for R5 .

          

 1 0 0 0 0 

0 1 0 0 0

          

 5
Solution: From (a),  6 , 3 , 0 , 1 , 0 form a basis for R .
        


 0 0 1 0 0 
 
3 0 0 0 1
 

          

 1 2 −1 0 3 
          
0 −1  3  1 −1

6. Let S =   ,   ,   ,   ,   and V = span(S). Find a subset

 1
 0 5 2 1 

3 1 12 5 4
 
S ′ ⊆ S such that S ′ forms a basis for V .

Solution:    
1 2 −1 0 3 1 0 5 2 1
0 −1 3 1 −1 RREF 0
  1 −3 −1 1

1 0 −−−−→  
5 2 1 0 0 0 0 0
3 1 12 5 4 0 0 0 0 0
     

 1 2 
    
0 −1

Let S ′ =  ,  .
1  0 

 
3 1
 
Extra problems
1. Suppose A and B are two matrices such that AB = 0. Show that the column space of
B is contained in the nullspace of A.


Solution: Write B = b1 b2 · · · bn , where bi is the i-th column of B. Then



0 = AB = A b1 b2 · · · bn = Ab1 Ab2 · · · Abn

By comparing the columns, we conclude that Abi = 0 for all i = 1, ..., n. Hence,
bi ∈ Null(A) for all i = 1, ..., n, that is, {b1 , ..., bn } ⊆ Null(A). Since the nullspace
of A is a subspace space and is thus closed under linear combinations, this shows that
Col(B) = span{b1 , ..., bn } ⊆ Null(A).

2. Let A be a n × m matrix and P an n × n matrix.

(a) If P is invertible, show that rank(PA) = rank(A).
Solution: Since P is invertible, we can write P = Ek · · · E1 , for some elementary
matrices E1 , ..., Ek . This means that PA = Ek · · · E1 A, which shows that PA is row
equivalent to A, and therefore they have the same row space, Row(PA) = Row(A).
Thus,
rank(PA) = dim(Row(PA)) = dim(Row(A)) = rank(A)
Alternative Solution. It is a fact that for an a by b matrix X and a b by c matrix Y,
we have
rank(XY) ≤ rank(Y). (*)
Let B = PA. Since P is invertible, we have P−1 B = A. Using the above fact (*),
we get
rank(A) = rank(P−1 B) ≤ rank(B) = rank(PA) ≤ rank(A).
Hence all the inequalities are equalities and we get rank(PA) ≤ rank(A).

(b) Given an example such that rank(PA) < rank(A).

Solution: Let A = In the n × n identity matrix and P = 0n the n × n zero matrix,
for some n ≥ 1. Then

rank(PA) = 0 < rank(A) = n.

(c) If rank(PA) = rank(A). Can we conclude that P is invertible? Justify your answer.
 
1 0 0
Solution: No. For example, let P = A = 0 1 0. Then PA = A and so
0 0 0
rank(PA) = rank(A).



3. Let A = a1 a2 · · · an and B = b1 b2 · · · bn be row equivalent m×n matrices,
where ai and bi are the i-th column of A and B, respectively, for i = 1, ..., n. Show that
for any c1 , c2 , ..., cn ∈ R,
c1 a1 + c2 a2 + · · · + cn an = 0
if and only if
c1 b1 + c2 b2 + · · · + cn bn = 0.

Solution: Since A and B are row equivalent, we can write B = Ek · · · E1 A = PA,

where P = Ek · · · E1 , for some elementary matrices E1 , ..., Ek . Then P is invertible.
Moreover,




b1 b2 · · · bn = B = PA = P a1 a2 · · · an = Pa1 Pa2 · · · Pan ,
−1
which shows that bi = Pai for i =1, 2, ..., n. Since P is invertible, we have ai = P bi
c1
 c2 
for i = 1, 2, ..., n. We set c =  .. . Then
 
.
cn
 
c1

 c2 

c1 a1 + c2 a2 + · · · + cn an = a1 a2 · · · an  ..  = Ac,

.
cn
 
c1

 c2 

c1 b1 + c2 b2 + · · · + cn bn = b1 b2 · · · bn  ..  = Bc = PAc.

.
cn

We have

c1 a1 + c2 a2 + · · · + cn an = 0
⇔ Ac = 0
⇔ PAc = P0 (because P is invertible.)
⇔ Bc = 0
⇔ c1 b1 + c2 b2 + · · · + cn bn = 0.

4. Let S = {v1 , v2 , . . . , vn } be a basis for a vector space V . Let u be a vector in V and let
c be a scalar. Prove the following:

(a) [u + v]S = [u]S + [v]S .

 Solution: We first write u and v in terms of the basis vectors, say

u = c1 v1 + c2 v2 + . . . + cn vn and v = d1 v1 + d2 v2 + . . . + dn vn .

Then
u + v = (c1 + d1 )v1 + (c2 + d2 )v2 + . . . + (cn + dn )vn
which implies
     
c1 + d 1 c1 d1
 c2 + d 2   c2   d 2 
[u + v]S =  ..  =  ..  +  ..  = [u]S + [v]S .
     
 .  . .
cn + d n cn dn

(b) [cu]S = c [u]S .

Solution: Similarly,
  
cc1 c1
 cc2   c2 
cu = (cc1 )v1 + (cc2 )v2 + . . . + (ccn )vn ⇒ [cu]S =  ..  = c  ..  = c [u]S .
   
 .  .
ccn cn

(c) Suppose u1 , u2 , . . . , uk are vectors in V . Note that for each i = 1, 2, . . . , k, [ui ]S is a

vector in Rn . By induction and using (a) and (b), it follows that if c1 , c2 , . . . , ck ∈ R,
then
[c1 u1 + c2 u2 + . . . + ck uk ]S = c1 [u1 ]S + c2 [u2 ]S + . . . + ck [uk ]S .
Prove that {u1 , u2 , . . . , uk } is linearly independent in V if and only if {[u1 ]S ,
[u2 ]S , . . . , [uk ]S } is linearly independent in Rn .

Solution: Assume that {u1 , u2 , . . . , uk } is linearly independent in V . Consider the

equation c1 [u1 ]S + . . . + ck [uk ]S = 0 which is a vector equation in Rn . By part
(c), the equation above can be rewritten as [c1 u1 + c2 u2 + . . . + ck uk ]S = 0. So the
coordinates of c1 u1 + . . . + ck uk with respect to the basis S are all zero, that is,
c1 u1 +. . .+ck uk = 0v1 +0v2 +. . .+0vn = 0. As {u1 , . . . uk } is linearly independent,
the equation above implies c1 = c2 = . . . = ck = 0, so {(u1 )S , . . . , (uk )S } is linearly
independent in Rn .
Conversely, assume {[u1 ]S , . . . , [uk ]S } is a linearly independent set in Rn . Consider
the equation
c1 u1 + c2 u2 + . . . + ck uk = 0
which implies
[c1 u1 + c2 u2 + . . . + ck uk ]S = [0]S = 0,
Using the result in part (c), we have

c1 [u1 ]S + . . . + ck [uk ]S = 0

and this implies that c1 = c2 = . . . = ck = 0 because {[u1 ]S , . . . , [uk ]S } is linearly

independent. Thus we have shown that {u1 , . . . , uk } is a linearly independent set
in V .