MTH102 : Linear Algebra

Assignment-04 (August-December 2024)

 
1 5
1. Given an example of two 2 by 2 matrices B and C such that B ̸= C but AB = AC, where A = .
3 15

Solution: Consider the following matrices:

   
5 0 0 5
B= and C= .
−1 0 0 −1

Clearly, B ̸= C, but  
0 0
AB = BA = .
0 0

Page 1 of 30
2. If the inverse of A2 is B, show that the inverse of A is AB. (Thus A is invertible whenever A2 is
invertible.)

Solution: We are given that A2 is invertible and its inverse is B, that is, A2 B = BA2 = I.
Then, it follows that
A(AB) = (AA)B = A2 B = I,
and

(AB)A = IABA
= (BA2 )ABA
= BA3 BA
= BA(A2 B)A
= BAIA
= BA2
= I.

Thus, we have seen that A(AB) = I and (AB)A = I. Hence, AB is the inverse of A.

Page 2 of 30
3. Find three 2 by 2 matrices, other than I and −I, that are their own inverses: A2 = I.

 
a b
Solution: If A = , then
c d
    2 
a b a b a + bc (a + d)b
A2 = = .
c d c d (a + d)c d2 + bc

In particular, take d = −a so that

 2   
a + bc 0 1 0
A2 = 2 = (a2 + bc) .
0 a + bc 0 1

There are many ways to choose a, b, c ∈ R such that they satisfy a2 + bc = 1. With any such choice
we will have A2 = I.
It is thus easy to see that the square of any of the following matrices is the identity matrix:
           
1 0 1 0 1 0 1 0 1 0 1 0
, , , , , ,...
0 −1 1 −1 2 −1 3 −1 4 −1 5 −1

Page 3 of 30
4. Give examples of 2 by 2 matrices matrices A and B such that
(a) A + B is not invertible although A and B are invertible.
(b) A + B is invertible although A and B are not invertible.
(c) All of A, B, and A + B are invertible.

Solution:

(a) Take A = I and B = −I. These matrices A and B are invertible. In fact, A−1 = A = I and
B −1 = B = −I. But, the matrix A + B = 0 is not invertible. Here 0 denotes the 2 by 2 zero
matrix.
   
1 0 0 0
(b) Take A = and B= .
0 0 0 1
Clearly, A and B are not invertible, but A + B = I is invertible.

(c) Take A = B = I so that A + B = 2I. Clearly, all of A, B, and A + B are invertible.

Page 4 of 30
5. Let A and B be n by n matrices such that all of A, B, and A + B are invertible. In this case, show that
C = A−1 + B −1 is also invertible, and find a formula for C −1 .

Solution:
Recall: The product of finitely many invertible matrices is invertible, with the inverse being the
product in the reverse order of the individual inverses. One can see it from a simple check up, for
example, for three invertible matrices, say, P , Q and R. We claim that P QR is invertible with its
inverse being R−1 Q−1 P −1 . In fact,

(P QR)(R−1 Q−1 P −1 ) = P Q(RR−1 )Q−1 P −1

= P QIQ−1 P −1
= P (QQ−1 )P −1
= P IP −1
= I,

and similarly one can check that (R−1 Q−1 P −1 )(P QR) = I.
This proves that R−1 Q−1 P −1 is the inverse of P QR.

Coming back to our main question, note that

A−1 (B + A)B −1 = A−1 BB −1 + A−1 AB −1

= A−1 I + IB −1
= A−1 + B −1 = C.

Thus, we see that C = A−1 + B −1 is the product of three invertible matrices, namely, A−1 , B + A
and B −1 . Therefore, the matrix C is invertible with

C −1 = B(B + A)−1 A.

Page 5 of 30
6. Under what conditions on their entries are A and B invertible?
   
a b c a b 0
A = d e 0 and B = c d 0 .
f 0 0 0 0 e

Solution:
(a) Let P13 denote the 3 by 3 permutation matrix which interchanges rows 1 and 3. We know that
P13 is invertible, which implies that A is invertible if and only if C = P13 A is invertible.
Note that
 
f 0 0
C = P13 A =  d e 0 .
a b c

So, we see that C is a lower triangular matrix, and we know that a triangular matrix (upper
or lower) is invertible if and only if every diagonal entry is non-zero.
We conclude that A is invertible if and only if all of c, e and f are non-zero.

(b) We claim that B is invertible if and only if (ad − bc)e ̸= 0.

(a) Case 1: If b = 0.
 
a 0 0
In this case, B =  c d 0 is a lower triangular matrix which is invertible if and only if
0 0 e
ade ̸= 0, which in this case is same as (ad − bc)e = ade ̸= 0.
(b) Case 2: If d = 0.
Let P12 denote the 3 by 3 permutation matrix which interchanges rows 1 and 2. We
know that P12 is invertible, which implies that B is invertible if and only if D = P12 A is
invertible.
Note that
 
c 0 0
D = P12 A = a b 0 .
0 0 e

Again, D being a lower triangular matrix, is invertible if and only if cbe ̸= 0, which in this
case is same as (ad − bc)e = −bce ̸= 0.
(c) Case 3: If bd ̸= 0.
Let E12 denote the elimination matrix corresponding to“replacing row 1 by the subtraction
1 −b/d 0
of (−b/d) times of row 2 from row 1”. That is, E12 = 0 1 0.
0 0 1
We know that E12 is invertible, which implies that B is invertible if and only if F = E12 B
is invertible.
Note that
a − bc
 
d
0 0
F = E12 B =  c d 0 .
0 0 e
bc


As earlier, F being a lower triangular matrix, is invertible if and only if a − d de ̸= 0,
which in this case is same as (ad − bc)e ̸= 0.

Page 6 of 30
7. (Remarkable) Let A and B be n by n matrices. Prove that I − BA is invertible if and only if I − AB
is invertible. [Hint: One can make use of the identity B(I − AB) = (I − BA)B.]

Solution: We shall prove that if I − AB is invertible, then I − AB is invertible. Since the role of
A and B is interchangeable, the other implication is automatic.
So, let us assume that I − AB is invertible. We shall show that (I − BA)−1 = I + B(I − AB)−1 A.
We shall make use of the following interesting identities:
B(I − AB) = (I − BA)B (both sides are equal to B − BAB) and A(I − BA) = (I − AB)A (both
sides are equal to A − ABA).
Note that

I = (I − BA) + BA
= (I − BA) + BIA
= (I − BA) + B (I − AB)(I − AB)−1 A

= (I − BA) + (B(I − AB)) (I − AB)−1 A

= (I − BA) + ((I − BA)B) (I − AB)−1 A

(using the identity B(I − AB) = (I − BA)B)
−1
= (I − BA) + (I − BA)B(I − AB) A
= (I − BA) I + B(I − AB)−1 A .

Repeating the similar calculations as above, we also have

I = (I − BA) + BA
= (I − BA) + BIA
= (I − BA) + B (I − AB)−1 (I − AB) A

= (I − BA) + B(I − AB)−1 ((I − AB)A)

= (I − BA) + B(I − AB)−1 (A(I − BA))

(using the identity A(I − BA) = (I − AB)A)
−1
= (I − BA) + B(I − AB) A(I − BA)
= I + B(I − AB)−1 A (I − BA).

We have shown that (I − BA) I + B(I − AB)−1 A = I + B(I − AB)−1 A (I − BA) = I, which

proves that I − BA is invertible and (I − BA)−1 = I + B(I − AB)−1 A.

Page 7 of 30
8. Invert these matrices A by the Gauss-Jordan method starting with [A I]:
 
  1 0 0 0
1 0 0 1
4
1 0 0
A = 2 1 3 and A=
1 1
.
3 3
1 0
0 0 1 1 1 1
2 2 2
1

Solution:
(a) We will find the inverse of the matrix
 
1 0 0
A = 2 1 3 .
0 0 1
To use the Gauss-Jordan method, we will write the matrix A and the identity matrix together
as follows:  
1 0 0 1 0 0
[ A | I ] = 2 1 3 0 1 0 .
0 0 1 0 0 1
Our goal is to transform the left side (the matrix A) into the identity matrix while performing
the same row operations on the right side (the identity matrix).
Step 1: We shall eliminate the 2 appearing in the (2, 1)th position by performing the operation
R2 → R2 − 2R1 . This gives us:
 
1 0 0 1 0 0
0 1 3 −2 1 0 .
0 0 1 0 0 1

Step 2: We shall eliminate the 3 appearing in the (2, 3)th position by performing the operation
R2 → R2 − 3R3 . This gives us:
 
1 0 0 1 0 0
0 1 0 −2 1 −3 .
0 0 1 0 0 1

Thus, the inverse of the matrix A is

 
1 0 0
A−1 = −2 1 −3 .
0 0 1

(b) Now, consider the matrix  

1 0 0 0
1 1 0 0
A= 4 .
1 1
1 0
3 3
1 1 1
2 2 2 1
As in part (a), we put together the matrix A and the identity matrix as follows:
 
1 0 0 0 1 0 0 0
1 1 0 0 0 1 0 0
[A | I ] =  4 .
1 1 1 0 0 0 1 0
3 3
1 1 1
2 2 2 1 0 0 0 1

Page 8 of 30
Our aim is to transform the left side matrix into the identity matrix while performing the same
operations on the right side.
Step 1: We first eliminate all the entries of the first column except the (1, 1)th entry by per-
forming the following operations one-by-one:
1 1 1
R2 → R2 − R1 , R3 → R3 − R1 and R4 → R4 − R1 .
4 3 2
This results in:  
1 0 0 0 1 0 0 0
0
 1 0 0 − 14 1 0 0.
1
0
3 1 0 − 31 0 1 0
1 1
0 2 2 1 − 21 0 0 1

Step 2: We now eliminate the entries below the 1 appearing in the (2, 2)th position by exercising
the following operations:
1 1
R3 → R3 − R2 and R4 → R4 − R2 .
3 2
This yields:  
1 0 0 0 1 0 0 0
0
 1 0 0 − 41 1 0 0
.
0 0 1 0 − 41 − 13 1 0
1
0 0 2 1 − 83 − 12 0 1

Step 3: Eliminate the 12 appearing in the (4, 3)th position by performing the operation R4 →
R4 − 21 R3 . This gives us:  
1 0 0 0 1 0 0 0
0 1 0 0 − 1 1 0 0
 4 .
0 0 1 0 − 1 − 1 1 0
4 3
0 0 0 1 − 41 − 13 − 21 1

Thus, the inverse of A is  

1 0 0 0
− 1 1 0 0
A−1 =
− 1
4 .
4 − 13 1 0
− 41 − 13 − 12 1

Page 9 of 30
9. True or false (with a counterexample if false and a reason if true):
(a) A 4 by 4 matrix with a row of zeros is not invertible.
(b) A matrix with 1s down the main diagonal is invertible.

Solution:
(a) The statement is TRUE.
Let A be any 4 by 4 matrix and i ∈ {1, 2, 3, 4} be arbitrary. Suppose that the ith row of A has
all the entries equal to zero. Let B be any 4 by 4 matrix. After multiplying the matrices A and
B, observe that the entry in the (i, i)th position of the matrix AB is zero. This implies that
AB is not the identity matrix. Since B was chosen arbitrarily, we conclude that the matrix A
is not invertible.
(b) The statement is FALSE.
Consider the matrix  
0 0 0
A = 1 0 0 .
1 1 0
Clearly, each entry of A below the main diagonal is 1, but A is not invertible.

Page 10 of 30
10. Which of the following subsets of R3 are actually subspaces?
(a) The collection of vectors b = (b1 , b2 , b3 ) with first component b1 = 0.
(b) The collection of vectors b = (b1 , b2 , b3 ) with first component b1 = 1.
(c) The collection of vectors b = (b1 , b2 , b3 ) with b2 b3 = 0 (this is, the union of two subspaces, the plane
b2 = 0 and the plane b3 = 0).
(d) All combinations of two given vectors (1, 1, 0) and (2, 0, 1).
(e) The plane of vectors b = (b1 , b2 , b3 ) that satisfy b3 − b2 + 3b1 = 0.

Solution:

(a) We are given the set W1 = {(b1 , b2 , b3 ) ∈ R3 : b1 = 0}. We can see that this set represents the
yz-plane in R3 .
Clearly, this set is non-empty, as the null vector (0, 0, 0) ∈ W1 .
Let (b1 , b2 , b3 ), (a1 , a2 , a3 ) ∈ W1 . So here b1 = 0 and a1 = 0, and this gives b1 + a1 = 0.
Therefore, (b1 , b2 , b3 ) + (a1 , a2 , a3 ) = (b1 + a1 , b2 + a2 , b3 + a3 ) ∈ W1 since b1 + a1 = 0.
Take any c ∈ R and (b1 , b2 , b3 ) ∈ W1 .
Then, c(b1 , b2 , b3 ) = (c b1 , c b2 , c b3 ) ∈ W1 since c b1 = 0.
Hence, W1 is a subspace of R3 .

(b) Here the given set is W2 = {(b1 , b2 , b3 ) ∈ R3 : b1 = 1}. Clearly, the null vector (0, 0, 0) ̸∈ W2 ,
and therefore W2 is not a subspace of R3 .
(c) The given set is W3 = {(b1 , b2 , b3 ) ∈ R3 : b2 b3 = 0}. We can see that this set is the union of two
subspaces, the plane b2 = 0 and the plane b3 = 0.
Take two vectors (1, 1, 0), (1, 0, 1) ∈ W3 .
Note that (1, 1, 0) + (1, 0, 1) = (1, 1, 1) ̸∈ W3 .
Hence, W3 is not a subspace of R3 .
(d) The given set is W4 = {s(1, 1, 0) + t(2, 0, 1) ∈ R3 : s, t ∈ R}.
Clearly, this set is non-empty, as the null vector (0, 0, 0) = 0(1, 1, 0) + 0(2, 0, 1) ∈ W4 .
Take any two vectors (b1 , b2 , b3 ), (a1 , a2 , a3 ) ∈ W4 . Then, there exists s1 , s2 , t1 , t2 ∈ R such that

(b1 , b2 , b3 ) = s1 (1, 1, 0) + t1 (2, 0, 1) and (a1 , a2 , a3 ) = s2 (1, 1, 0) + t2 (2, 0, 1).

Now,
(b1 , b2 , b3 ) + (a1 , a2 , a3 ) = (s1 + s2 )(1, 1, 0) + (t1 + t2 )(2, 0, 1),
and since s1 + s2 , t1 + t2 ∈ R, it follows that (b1 , b2 , b3 ) + (a1 , a2 , a3 ) ∈ W4 .
Also, for any c ∈ R, we have

c(b1 , b2 , b3 ) = (cs1 )(1, 1, 0) + (ct1 )(2, 0, 1),

and since cs1 , ct1 ∈ R, we get that c(b1 , b2 , b3 ) ∈ W4 .

Hence, W4 is a subspace of R3 .

Page 11 of 30
(e) The given set is W5 = {(b1 , b2 , b3 ) ∈ R3 : b3 − b2 + 3b1 = 0}.
Clearly, this set is non-empty, as the null vector (0, 0, 0) = 0(1, 1, 0) + 0(2, 0, 1) ∈ W5 .
Let (b1 , b2 , b3 ), (a1 , a2 , a3 ) ∈ W5 be any two vectors.
Then, we have b3 − b2 + 3b1 = 0 and a3 − a2 + 3a1 = 0, which implies that

(b3 + a3 ) − (b2 + a2 ) + 3(b1 + a1 ) = 0.

Therefore, (b1 , b2 , b3 ) + (a1 , a2 , a3 ) = (b3 + a3 , b2 + a2 , b1 + a1 ) ∈ W5 .

For any c ∈ R, we have b3 − b2 + 3b1 = 0 =⇒ c b3 − c b2 + 3c b1 = c(b3 − b2 + 3b1 ) = 0, so that

c(b1 , b2 , b3 ) = (c b1 , c b2 , c b3 ) ∈ W5 .

Hence, W5 is a subspace of R3 .

Page 12 of 30
11. Describe the column space and the nullspace of the following matrices:
     
1 −1 0 0 3 0 0 0
A= and B= and C= .
0 0 1 2 3 1 2 3

Solution: We know that the column space C(A) of an m × n matrix A is the collection of all linear
combinations of the columns of A, and that C(A) is a subspace of Rm .
On the other hand, the null space N (A) of an m × n matrix A is the collection of all vectors v ∈ Rn
such that Av = 0, and that N (A) is a subspace of Rn .
 
1 −1
• Given matrix A = .
0 0
We have
     
1 −1
C(A) = s +t : s, t ∈ R
0 0
   
1
= (s − t) : s, t ∈ R
0
   
1
= α :α∈R .
0
 
1
Clearly, C(A) is collection of all scalar multiples of , that is, C(A) is the x-axis in R2 .
0
          
x 1 −1 x 0 x−y 0
On the other hand, for v = , we have Av = = ⇐⇒ = .
y 0 0 y 0 0 0
Comparing the entries of the last two columns we get x − y = 0 ⇐⇒ x = y.
So, N (A) = {(x, x) : x ∈ R} = {x(1, 1) : x ∈ R}.
Thus, N (A) is the collection of all scalar multiples of (1, 1), that is, N (A) is the straight line
in R2 passing through (1, 1) and (0, 0).
 
0 0 3
• Given matrix B = .
1 2 3
Observe that the first two columns are scalar multiple of each other, as we have (0, 2) = 2(0, 1)
and (0, 1) = 21 (0, 2). So these two columns together are linearly dependent columns, and while
taking linear combinations, we can drop any one of these two.
 
0 3
Thus, C(B) = C(D) where D = .
1 3
One can easily check that the matrix D is invertible, which implies that C(D) = R2 , and hence
C(B) = R2 .
 
x
On the other hand, for v = y , we have
z
 
  x      
0 0 3   0 3z 0
Bv = y = ⇐⇒ = .
1 2 3 0 x + 2y + 3z 0
z

Page 13 of 30
 The above system of linear equations is same as z = 0 and x + 2y = 0.
Thus,

N (B) = {(x, y, z) ∈ R3 : x + 2y = 0, z = 0}
= {(x, y, 0) ∈ R3 : x + 2y = 0}
= {(−2y, y, 0) : y ∈ R}
= {y(−2, 1, 0) : y ∈ R}.

Hence, the null space N (B) of B is the straight line in R3 which lies in the xy-plane and passes
through (0, 0, 0) and (−2, 1, 0).
 
0 0 0
• Given matrix C = .
1 2 3
Here observe that each column is the scalar multiple of every other column. So, while taking
linear combinations, we can choose any one of these.
   
0
Thus, the column space of C is C(C) = α :α∈R .
1
That is, the column space of the matrix C is the straight line in R2 passing through (0, 0) and
(0, 1).
 
x
On the other hand, for v = y , we have
z
 
  x      
0 0 0   0 0 0
Cv = y = ⇐⇒ =
1 2 3 0 x + 2y + 3z 0
z

The above system of linear equations is same as x + 2y + 3z = 0.

Thus,

N (A) = {(x, y, z) ∈ R3 : x + 2y + 3z = 0}
= {(−2y − 3z, y, z) ∈ R3 : y, z ∈ R}
= {y(−2, 1, 0) + z(−3, 0, 1) : y, z ∈ R}.

Hence, the null space N (C) of the matrix C is the plane in R3 passing through (0, 0, 0),
(−2, 1, 0) and (−3, 0, 1).

Page 14 of 30
12. Which of the following descriptions are correct? The solutions x of
 
 x
1  1
  
1 1 0
Ax = x2 =
1 0 2 0
x3

form

(a) a plane;
(b) a line;
(c) a point;
(d) a subspace;
(e) the nullspace of A;
(f) the column space of A.

Solution: We are given the system of linear equations governed by

 
  x1  
1 1 1   0
Ax = x2 = .
1 0 2 0
x3

By definition, the null space N (A) is the collection of all vectors x ∈ R3 such that Ax = 0. Thus,
the collection of solutions x of Ax = 0 is the null space of A.

Since N (A) is a subspace of R3 , the solutions x of Ax = 0 form a subspace of R3 .

However, it is not the column space of A, because the column space C(A) is the collection of vectors
b ∈ R2 such that Ax = b for some x ∈ R3 .

Now, solving Ax = 0, we get the system of equations:

x1 + x2 + x3 = 0,
x1 + 0x2 + 2x3 = 0.

From the second equation, we have

x1 = −2x3

Substituting this into the first equation, we get

−2x3 + x2 + x3 = 0 ⇐⇒ x2 − x3 = 0 ⇐⇒ x2 = x3 .
 
x1
Thus, x = x2  solves Ax = 0 if and only if
x3
     
x1 −2t −2
x2  =  t  = t  1  , for some t ∈ R.
x3 t 1

Page 15 of 30
Therefore, the solution space, which is N (A), is given by
   
 −2 
N (A) = t  1  : t ∈ R .
1
 

   
0 −2
This represents a line passing through the points 0 and  1  .
0 1

Conclusion: The solutions x of Ax = 0

(a) do not form a plane;

(b) form a line;

(c) is not just a point;

(d) form a subspace;
(e) is the null space of A;

(f) is not the column space of A.

Page 16 of 30
13. The columns of AB are combinations of the columns of A. This means: The column space of AB is
contained in (possibly equal to) the column space of A. Give an example where the column spaces of A
and AB are not equal.

Solution:

(a) Let A be an m × n matrix and B be an n × l matrix. We want to show that

C(AB) ⊆ C(A).

Let b ∈ C(AB). Then, by definition, there exists x ∈ Rl such that

b = (AB)x.

Thus, if we write v = Bx then we have b = Av which implies that b ∈ C(A).

Hence, C(AB) ⊆ C(A).

(b) To find an example where the column spaces of A and AB are not equal, consider the following
matrices A and B:    
1 0 0 0
A= , B= .
0 0 1 0
so that     
1 0 0 0 0 0
AB = = .
0 0 1 0 0 0

Note that the column space C(A) of A is the x-axis in R2 , whereas the column space C(AB) of
AB is the trivial subspace {0} of R2 . Thus, in this particular example, the column space of AB
is a proper subspace of the column space of A.

Page 17 of 30
14. True or false (with a counterexample if false)?
(a) The vectors b that are not in the column space C(A) form a subspace.
(b) If C(A) contains only the zero vector, then A is the zero matrix.
(c) The column space of 2A equals the column space of A.
(d) The column space of A − I equals the column space of A.

Solution: Here, we prove that the statements (a) and (d) are false, and the statements (b) and (c)
are true.
Recall that given an m by n matrix A, the column space C(A) of A is the vector subspace of Rm
consisting precisely of all possible linear combinations of columns Cj , 1 ≤ j ≤ n of A. Also, we know
that b ∈ C(A) if and only if there exists an x ∈ Rn such that b = Ax.
We denote the zero vector in Rm by 0.

(a) The statement is FALSE.

 
Let A = 1 , which is a 1 by 1 matrix. Note that C(A) = R, and hence, the set of vectors that
are not in the column space of A being empty, is not a subspace of R.
In fact, we can prove that the vectors that are not in the column space of A never form a
subspace of Rm , for any m by n matrix A.
Let A be an m by n matrix with columns Cj , 1 ≤ j ≤ n. Consider V to be the set of all vectors
b that are not in C(A), that is,

V = {b ∈ Rm : b ∈
/ C(A)}.

Case 1: The set V is empty.

In this case, V is not a subspace of Rm , as a subspace has to be nonempty by its very definition.
Case 2: The set V is nonempty.
Choose b ∈ V . We claim that −b ∈ V .
If −b ∈ / V , then we have that −b ∈ C(A). By definition of C(A), this implies that there exist
b1 , b2 , . . . , bn ∈ R such that −b = b1 C1 + b2 C2 + · · · + bn Cn . Hence, we get b = (−b1 )C1 +
(−b2 )C2 + · · · + (−bn )Cn , showing that b ∈ C(A), which contradicts the hypothesis that b ∈ V .
Therefore, we conclude that −b ∈ V .
If V is a subspace of Rm , then we should have b + (−b) = 0 ∈ V . But note that 0 = 0C1 +
0C2 + · · · + 0Cn , and so 0 ∈ C(A), that is, 0 ∈
/ V . Therefore, V does not form a subspace of
Rm .
(b) The statement is TRUE.
Let A be an m by n matrix such that C(A) = {0}. This says that any linear combination
P of the
columns of A is equal to 0. In particular, for each 1 ≤ j ≤ n, we have Cj = 1Cj + i̸=j 0Ci = 0.
That is, every column of A is the zero vector. Hence, A is the zero matrix.
(c) The statement is TRUE.
Let A be an m by n matrix. We will show that C(A) = C(2A).

Page 18 of 30
 n
Let b ∈ C(A) be

arbitrary. This means ′that1 there nexists an x ∈ R with Ax = b, and hence, we
have (2A) 2 x = b. Thus, the vector x = 2 x ∈ R is such that (2A)x′ = b, which implies that
1

b ∈ C(2A). This shows that C(A) ⊆ C(2A).

On the other hand, if c ∈ Rm is such that c ∈ C(2A), then there exists a y ∈ Rn such that
(2A)y = c, which implies that A(2y) = c. In other words, the vector y ′ = 2y ∈ Rn is such that
Ay ′ = c. Hence, it follows that c ∈ C(A). This proves that C(2A) ⊆ C(A).
Combining both of the above inclusions, we conclude that the column space of A equals the
column space of 2A.
(d) The statement is FALSE.
Fix any positive integer m and consider A = I, the m by m identity matrix. We know that the
column space of any m by m invertible matrix is Rm . In particular, the column space C(I) of
I, is the whole space Rm . But note that A − I = I − I = 0m×m , the m by m zero matrix, and
clearly, the column space of the zero matrix is {0}. Thus, the column space of A is not equal
to the column space of A − I here.

Page 19 of 30
15. Reduce A to echelon form, to find its rank. Which variables are free?
 
1 2 0 1
A = 0 1 1 0 .
1 2 0 1

Solution: We have
 
1 2 0 1
A = 0 1 1 0 .
1 2 0 1

Subtracting first row from third row we get

 
1 2 0 1
0 1 1 0 ,
0 0 0 0

which is the row echelon form of the matrix A.

Note that the first two

 columns
 contain pivots, namely a11 = 1 and a22 = 1. If we write a general
x1
x2 
vector in R4 by x = x3 , then x1 , x2 are the pivot variables and x3 , x4 are the free variables of the


x4
system Ax = 0.

The rank of the matrix A is equal to the number of pivot variables in its echelon form. Therefore,
the rank of A is 2.

Page 20 of 30
16. Under what conditions on b1 and b2 (if any) does Ax = b have a solution?
   
1 2 0 3 b1
A= , b= .
2 4 0 7 b2
Find two vectors in the nullspace of A, and the complete solution to Ax = b.

   
1 2 0 3 b1
Solution: We are given A = & .
2 4 0 7 b2
Let us convert A into a row echelon form, thus changing [A b] to [U c].

   
1 2 0 3 b1 Row27→Row2−2Row1 1 2 0 3 b1
[A b] = −−−−−−−−−−−−−−→ = [U c] .
2 4 0 7 b2 0 0 0 1 b2 − 2b1
 
1 2 03
Note that in U = , there are two pivot entries, namely a11 = 1 and a24 = 1. Thus, if
0 0 01
 
x1
 x2 
we write a general vector in R4 by x = x3 , then x1 , x4 are the pivot variables and x2 , x3 are the


x4
free variables of the system Ax = U x = 0.

Recall that in an m × n row echelon matrix U , if the number of pivot entries is m then the column
space of the corresponding matrix A is equal to Rn . Thus, in our case, Ax = b is solvable for all
b ∈ R2 .

Furthermore, one can compute a particular solution xp of U x = c, by putting all free variables to
be 0. In our case, it leads to the choice of x2 = x3 = 0 and solving
 
  x1  
1 2 0 3  0=
 b1
.
0 0 0 1 0 b2 − 2b1
x4
The above system is nothing but
x1 + 3x4 = b1 and x4 = b2 − 2b1 ,
implying
xp = (7b1 − 3b2 , 0, 0, b2 − 2b1 ) .

In order to get complete solutions, we look at the nullspace of A which is same as the nullspace of
U . That is, we solve U x = 0. In doing so, we solve the system for as many number of times as the
number of free variables, by “choosing exactly one free variable 1 and all others 0” each time. In
the present case, x2 and x3 are free variables.
The case of x2 = 1 and x3 = 0: Let us solve
 
  x1  
1 2 0 3 
1= 0 .

U xn =
0 0 0 1 0 0
x4

Page 21 of 30
The above system is nothing but

x1 + 2 + 3x4 = 0 and x4 = 0,

implying
xn = (−2, 1, 0, 0) .

The case of x2 = 0 and x3 = 1: Let us solve

 
  x1  
1 2 0 0= 0 .
3  
U xn =
0 0 0 1 1 0
x4

The above system is nothing but

x1 + 3x4 = 0 and x4 = 0,

implying
xn = (0, 0, 1, 0) .

Hence,
N (A) = {α (−2, 1, 0, 0) + β (0, 0, 1, 0) : α, β ∈ R} ,
and the complete set of solutions to Ax = b is equal to

{xp + α (−2, 1, 0, 0) + β (0, 0, 1, 0) : α, β ∈ R} ,

where xp = (7b1 − 3b2 , 0, 0, b2 − 2b1 ) .

Page 22 of 30
17. Write a 2 by 2 system Ax = b with many solutions xn but no solution xp . (Therefore the system has no
solution.) Which b’s allow an xp ?

   
1 1 b
Solution: Let us consider the 2 × 2 matrix A = and an arbitrary vector b = 1 ∈ R2 .
1 1 b2
Let us convert A into a row echelon form, thus changing [A b] to [U c].

   
1 1 b1 Row27→Row2−Row1 1 1 b1
[A b] = −−−−−−−−−−−−−→ = [U c] .
1 1 b2 0 0 b2 − b1

It is now clear that if we take b1 ̸= b2 then there does not exist any solution xp of Axp = b
( ⇐⇒ U xp = c).
 
1 1
Observe that in U = , there is one pivot entry, namely a11 = 1, and thus if we write a general
 0 0
x
vector in R2 by x = 1 , then x1 is the pivot variable whereas x2 is the free variable for the system
x2
Ax = U x = 0.
Now, let us solve U x = 0. In doing so, we solve the system for as many number of times as the
number of free variables, by “choosing exactly one free variable 1 and all others 0” each time. In
the present case, we have only one free variable x2 , so we choose x2 = 1 and solve
    
1 1 x1 0
U xn = = .
0 0 1 0

The above system is nothing but

x1 + 1 = 0
implying
xn = (−1, 1).

Thus, we have infinitely many xn ’s, namely every member of the nullspace N (A) which is the line
passing through (−1, 1) and (0, 0).
   
b1 b
Coming back to the possibilities of xp , if we take b1 = b2 , that is, b = or equivalently c = 1 ,
b1 0
then we can solve U xp = c by choosing the free variable x2 to be 0. More precisely, we have to solve
    
1 1 x1 b
U xp = = 1 ,
0 0 0 0

which is same as x1 = b1 , implying

xp = (b1 , 1).

Page 23 of 30
18. Find the reduced row echelon forms R and the rank of these matrices:
(a) The 3 by 4 matrix of all 1s.
(b) The 4 by 4 matrix with aij = (−1)ij .
(c) The 3 by 4 matrix with aij = (−1)j .

Solution: Let us abbreviate “reduced row echelon form” by “RREF”.

(a) The given matrix is:

   
1 1 1 1 1 1 1 1
Row27→Row2−Row1
A = 1 1 1 1 −−−−−−−−−−−−−→ 0 0 0 0 (RREF )
Row37→Row3−Row1
1 1 1 1 0 0 0 0

It is evident from the reduced row echelon form that the number of pivot variables for the matrix
is exactly one, namely the pivot entry is at a11 (see page 75 of Lecture 12). Therefore, the rank
of the given matrix is 1.
(b) The given matrix is:

(−1)1·1 (−1)1·2 (−1)1·3 (−1)1·4

 
(−1)2·1 (−1)2·2 (−1)2·3 (−1)2·4 
A=
(−1)3·1

(−1)3·2 (−1)3·3 (−1)3·4 
(−1)4·1 (−1)4·2 (−1)4·3 (−1)4·4
 
−1 1 −1 1
 1 1 1 1
=
−1

1 −1 1
1 1 1 1
 
1 −1 1 −1
Row17→−Row1  1 1 1 1
−−−−−−−−−−−−−→  
Row47→Row4−Row2 −1 1 −1 1
0 0 0 0
 
1 −1 1 −1
Row27→Row2−Row1  0 2 0 2
−−−−−−−−−−−−−→  
Row37→Row3+Row1  0 0 0 0
0 0 0 0
 
1 0 1 0
Row17→Row1+ 21 Row2
 0 2 0 2
−−−−−−−−−−−−−−→   
0 0 0 0
0 0 0 0
 
1 0 1 0
Row27→ 12 Row2  0 1 0 1
−−−−−−−−−−→    (RREF).
0 0 0 0
0 0 0 0

The reduced row echelon form shows that the matrix has two pivot variables, namely the pivot
entries are at a11 and a22 (refer to page 75 of Lecture 12). Hence, the rank of the matrix is 2.

Page 24 of 30
(c) The given matrix is:

(−1)1 (−1)2 (−1)3 (−1)4

 

A = (−1)1 (−1)2 (−1)3 (−1)4 

(−1)1 (−1)2 (−1)3 (−1)4
 
−1 1 −1 1
= −1 1 −1 1
−1 1 −1 1
 
−1 1 −1 1
Row27→Row2−Row1
−−−−−−−−−−−−−→  0 0 0 0
Row37→Row3−Row1
0 0 0 0
 
1 −1 1 −1
Row17→−Row1
−−−−−−−−−−→ 0 0 0 0  (RREF )
0 0 0 0

From the reduced row echelon form, it is clear that the matrix has exactly one pivot variable,
namely the pivot entry is at a11 (see page 75 of Lecture 12). Therefore, the rank of the given
matrix is 1.

Page 25 of 30
19. Choose the number r so that (if possible) the ranks are (a) 1, (b) 2, (c) 3:
 
6 4 2
A = −3 −2 −1 .
9 6 r

 
6 4 2
Solution: The given matrix is: A = −3 −2 −1 .
9 6 r
Now, we try to compute the row echelon form of this matrix A.
 
6 4 2
A = −3 −2 −1
9 6 r
 
−3 −2 −1
Row1←→Row2
−−−−−−−−−−→  6 4 2
9 6 r
 
−3 −2 −1
Row27→Row2+2.Row1
−−−−−−−−−−−−−−→  0 0 0 
Row37→Row3+3.Row1
0 0 r−3
 
−3 −2 −1
Row2←→Row3
−−−−−−−−−−→  0 0 r − 3 .
0 0 0

The following are evident from the above row echelon form (refer to page 75 of Lecture 12) of A:

(a) If r − 3 = 0, then the number of pivot variables is exactly one, namely the pivot entry is at a11 .
Therefore, the rank of the given matrix is 1 if r = 3.

(b) If r ̸= 3, then the number of pivot variables is two, namely the pivot entries are at a11 and a23
and the rank is 2.
(c) The number of pivot variables can never be 3. Thus, the rank can never be 3 for any value of r.

Page 26 of 30
20. Reduce to U x = c (Gaussian elimination) and then Rx = d:
 
 x
3  1
  
1 0 2 2
x2
A x = 1 3 2 0 
x3 
 =  5  = b.
2 0 4 9 10
x4

Find a particular solution xp and all nullspace solutions xn .

Solution: The given matrix equation is Ax = b, that is,

 
 x
1 0 2 3  1
  
2
1 3 2 0 x2  =  5  . (1)
x3 
2 0 4 9 10
x4

In the following, for i ∈ {1, 2, 3}, Ri denotes the ith row.

• We will first reduce Eq. (1) to a matrix equation of the form U x = c, where U is an echelon
form, by performing row operations on the matrices A and b.
For that, we replace R2 and R3 by R2 − R1 and R3 − 2R1 respectively in Eq. (1), to get
 
 x
0 2 3  1
  
1 2
x2   
0 3 0 −3  x3  = 3 . (2)
0 0 0 3 6
x4

We have circled the pivot elements. Note that in the columns containing the pivot elements
(except that in R3 ), only zeros are there below each pivot element, and each pivot element lies
to the right of the pivot element in the row above. This says that the 3 by 4 matrix on the
left hand side of Eq. (2) is a required echelon form. Hence, Eq. (2) is of the form U x = c.
• We further reduce this equation to obtain a matrix equation of the form Rx = d, where R is
the reduced row echelon form, as follows.
Step 1: Replace R2 by 13 R2 and R3 by 31 R3 in Eq. (2), to make all the pivot elements 1. This
gives us the equation  
 x
2 3  1
  
1 0 2
x2   
0 1 0 −1 
x3  = 1 . (3)
0 0 0 1 2
x4

Step 2: Now we use the pivot elements in R2 and R3 to produce zeros above the pivot elements.
Observe that there is already a zero above the pivot element in R2 . So, we are left to change
the entries above the pivot element in R3 to zeros. For that, replace R2 by R2 + R3 , and R1
by R1 − 3R3 in Eq. (3). We get the equation
 
 x
0 2 0  1
  
1 −4
x2   
0 1 0 0  x3  = 3 . (4)
0 0 0 1 2
x4

Page 27 of 30
 Note that Eq. (4) is of the form Rx = d, where
   
1 0 2 0 −4
R = 0 1 0 0 and d =  3  .
0 0 0 1 2

• As the first, second and fourth columns have pivot elements, and the third column doesn’t
have a pivot element, so x1 , x2 and x4 are pivot variables, whereas x3 is the free variable.

• To find a particular solution xp of the equation Axp = b, we set the free variable x3 to zero in
Eq. (4). Then the matrix multiplication gives us that x1 = −4, x2 = 3 and x4 = 2. Hence, a
particular solution of the given equation Ax = b is
 
−4
3
xp = 
 0 .


• Now to determine all nullspace solutions, we first find “special solutions” (whose all possible
linear combinations give all nullspace solutions) of Ax = 0 (or U x = 0 or Rx = 0) by setting
the free variable x3 to 1, and solving the equation Rx = 0 for the pivot variables, that is, we
solve the following equation.
 
 x
1 0 2 0  1
  
0
0 1 0 0 x2  = 0 .
1
0 0 0 1 0
x4

After solving the above equation, we obtain that x1 = −2, x2 = 0 and x4 = 0. Thus, all
nullspace solutions of the equation Ax = b are of the form
 
−2
0
xn = t  1 ,


where t is a real number.

Page 28 of 30
      
1 1 0
21. If the complete solution to A x = is x = +c , then find A.
3 0 1

     
1 0 1
Solution: The equation A +c = is meaningful if and only if A is a 2 by 2 matrix.
0 1 3
 
a11 a12
• Let A = , where a11 , a12 , a21 and a22 are arbitrary real numbers.
a21 a22

Suppose the complete solution to the equation

 
1
Ax = (5)
3
   
1 0
is x = +c , where c is a real number.
0 1
Then,aparticular solution xp can be given by putting c = 0 in the complete solution. Thus,
1
xp = solves Eq. (5), that is, we have
0
    
a11 a12 1 1
= .
a21 a22 0 3

This implies that a11 = 1 and a21 = 3.

 
0
On the other hand, it follows from the given set of solutions that and all its scalar multiples
1
form the nullspace of A. Thus, we have
    
1 a12 0 0
= ,
3 a22 1 0

from which we get that a12 = 0 and a22 = 0.

 
1 0
Hence, we have A = .
3 0
 
1 0
• We will now show that if A = , then the complete solution of the equation
3 0
    
1 0 x1 1
= (6)
3 0 x2 3
     
x1 1 0
is x = = +c .
x2 0 1

Replacing row 2 in Eq. (6) by “the subtraction of 3 times of row 1 from row 2” gives us the
equation     
1 0 x1 1
= , (7)
0 0 x2 0
which holds if and only if x1 = 1.

Page 29 of 30
  
1
Thus, the solutions of Eq. (6) (or, equivalently Eq. (7)) are precisely of the form , where c
c
     
x1 1 0
is a real number. In other words, the complete solution to Eq. (6) is x = = +c .
x2 0 1
     
1 1 0
The above arguments show that the complete solution to the equation Ax = is x = +c
3 0 1
 
1 0
if and only if A = .
3 0

Page 30 of 30