MA 110:

Lecture 07

Saurav Bhaumik
Department of Mathematics
IIT Bombay

Spring 2025

Saurav Bhaumik, IIT Bombay MA 110: Lecture 07

Review of last lecture
We have discussed the following important notions.
Rank of a matrix
Vector subspaces (of Rn×1 ).
Basis of a subspace
Dimension of a subspace
Span of a subset of a vector subspace.
Null space and the column space of a matrix.
And we proved several important results such as:
Characterizations of a basis of a vector subspace
Rank-Nullity Theorem
Fundamental Theorem for Linear Systems:
Remark: The notion of a vector subspace of R1×m is defined
similarly. The corresponding notions of basis, dimension, span,
etc. are defined in an identical manner.
Saurav Bhaumik, IIT Bombay MA 110: Lecture 07
Let us restate two earlier results which are in conformity with
the Rank-Nullity Theorem. Let A be an n × n matrix. Then
A is invertible ⇐⇒ nullity A = 0 ⇐⇒ rank A = n .
Further, rank A = n ⇐⇒ C(A) = Rn×1 .

We are now in a position to state and prove a comprehensive

result regarding solutions of a system of m linear equations in
n unknowns that we started with, namely
a11 x1 + a12 x2 + · · · + a1n xn = b1 (1)
a21 x1 + a22 x2 + · · · + a2n xn = b2 (2)
.. .. .. .. ..
. . . . .
am1 x1 + am2 x2 + · · · + amn xn = bm (m)

As usual, we write this as Ax = b, where

T  T
A := [ajk ] ∈ Rm×n , x := x1 · · · xn and b := b1 · · · bm .
Saurav Bhaumik, IIT Bombay MA 110: Lecture 07
Theorem (Fundamental Theorem for Linear Systems: FTLS)
Let m, n ∈ N and A be an m × n matrix with real entries.
Suppose rank A = r .
(i) Homogeneous Linear System : Ax = 0 (H)
The solution space {x ∈ Rn×1 : Ax = 0} of (H) is a subspace
of Rn×1 of dimension n − r .
In particular, r = n if and only if 0 is the only solution of (H).
If r < n, then there are linearly independent solutions
x1 , . . . , xn−r of (H) and every solution of (H) is a unique linear
combination of these x1 , . . . , xn−r .

(ii) General Linear System : Ax = b with b ∈ Rm×1 (G)

(G) has a solution if and only if rank[A|b] = r . In this case, let

x0 be a particular solution of (G). If x is a solution of (G),
then x = x0 + xh , where xh is a solution of (H) above.
Saurav Bhaumik, IIT Bombay MA 110: Lecture 07
Proof. (i) The solution space of the homogeneous linear
system (H) is just the nullspace N (A) of A, and we have seen
that its dimension, that is, the nullity of A, is equal to n − r .
We note that r = n ⇐⇒ n − r = 0, that is, the dimension of
N (A) is zero; in other words, N (A) = {0}. This means that
0 is the only solution of (H).
Let now r < n. Then N (A) has a basis consisting of n − r
elements, say x1 , . . . , xn−r . Hence every element of the
solution space is a unique linear combination of the elements
in this basis.
(ii) Let b ∈ Rn×1 . Let c1 , . . . , cn be the n columns of A. Then
T
∈ Rn×1 .

Ax = x1 c1 + · · · + xn cn for x := x1 · · · xn

Hence Ax = b for some x ∈ Rn×1 if and only if b is a linear

combination of the columns of A, that is, b ∈ C(A).
Saurav Bhaumik, IIT Bombay MA 110: Lecture 07
Since every column of A is also a column of the augmented
matrix [A|b], the column space C(A) of A is contained in the
column space C([A|b]) of [A|b]. It follows that b ∈ C(A) if
and only if C([A|b]) = C(A), that is, the column rank of [A|b]
is equal to the column rank of A. So rank[A|b] = rank A = r .
Let x0 be a particular solution of (G), that is, let x0 ∈ Rn×1
satisfy Ax0 = b. Then for any x ∈ Rn×1 , we see that Ax = b
if and only if A(x − x0 ) = Ax − Ax0 = b − b = 0, that is, x is
a solution of (G) if and only if xh := x − x0 is a solution of (H).
The proof is complete.
Remark
The above theorem is of immense theoretical importance. It
tells us precisely when solutions exist, and also describes the
nature of solutions of a linear system of equations.
For example, it says that when there is a nonzero solution of a
homogeneous linear system, there are infinitely many solutions.
Saurav Bhaumik, IIT Bombay MA 110: Lecture 07
Further, when a homogeneous system has infinitely many
solutions, it says that they can be described by a one
parameter family, or a two parameter family etc.
It may seem that to implement the results of the above
theorem, we must first find the rank of the coefficient matrix
A of the linear system. This is not necessary.
We have already seen that we may directly proceed to find the
solutions of the linear system by considering the augmented
matrix [A|b] and transform the coefficient matrix A to a row
echelon form by the Gauss Elimination Method and then use
Back Substitution . This process itself reveals all possibilities.
In particular, when the rank r of A is less than the number n
of variables, we have shown how to construct a set S of basic
solutions of an homogeneous linear system. This set S is in
fact a basis of the solution space of the system. That is the
reason for using the terminology ‘ basic solutions ’.
Saurav Bhaumik, IIT Bombay MA 110: Lecture 07
Row Space and Column Space

Definition
Let A ∈ Rm×n . The row space of A, denoted R(A), is defined
as the subspace of R1×n spanned by the row vectors of A.

Let A ∈ Rm×n . Here are some important observations.

The row-rank of A is precisely the dimension of R(A).
If A′ ∈ Rm×n is obtained from A by an elementary row
operation, then R(A) = R(A′ ).
If A′ ∈ Rm×n is in REF, then the nonzero rows of A′ form
a basis of R(A′ ).
A basis of R(A) is given by the nonzero rows of its REF.
If A′ is obtained from A by an elementary row operation,
then C(A) need not be equal to C(A′ ).

Saurav Bhaumik, IIT Bombay MA 110: Lecture 07

Lemma
The columns of A corresponding to the pivotal columns of its
REF form a basis of C(A).

Proof. Let A′ be an REF of A. Suppose the columns

c′i1 , · · · , c′ir of A′ are the pivotal columns. Let ci1 , · · · , cir be
the corresponding columns of A. It is enough to prove that
ci1 , · · · , cir are linearly independent. If they were not so, then
there would be constants αi1 , · · · , αir , not all zero such that
αi1 ci1 + · · · + αir cir = 0.
Consider the vector x = (x1 , . . . , xn )T such that xj = αj if j is
one of i1 , . . . , ir , and xj = 0 other wise. Then Ax = 0, thus
A′ x = 0 as well. But this means,
αi1 c′i1 + · · · + αir c′ir = 0,
which is a contradiction because c′i1 , · · · , c′ir being the pivotal
columns of a matrix in REF, are linearly independent.
Saurav Bhaumik, IIT Bombay MA 110: Lecture 07
 Example: Consider the 5 × 6 matrix A and its REF A′ given by
   
1 3 −2 0 2 0 1 3 −2 0 2 0
2 6 −5 −2 4 −3 
 0 0 −1
 −2 0 −3
′
 
A= 0 0 0 0 0 0 , A =  0 0 0 0 0 6


0 0 5 10 0 15  0 0 0 0 0 0
2 6 0 8 4 18 0 0 0 0 0 0
Then rank A = 3. A basis of the row space R(A) is given by
     
1 3 −2 0 2 0 , 0 0 −1 −2 0 −3 , 0 0 0 0 0 6
whereas a basis for the column space C(A) is given by
     

 1 −2 0 
     
2 −5 −3
 

0 ,  0  ,  0  .
     


 0  5   15  
 
2 0 18
 

Saurav Bhaumik, IIT Bombay MA 110: Lecture 07

Determinants
You already know formulas for determinants of 1 × 1, 2 × 2
and 3 × 3 matrices. Let us recall them.
 
  a1 b1
det a1 = a1 , det = a1 b2 − a2 b1 and
a2 b2
 
a1 b1 c1
det a2 b2 c2  = a1 (b2 c3 −b3 c2 )−b1 (a2 c3 −a3 c2 )+c1 (a2 b3 −a3 b2 ),
a3 b3 c3
which is also equal to
a1 (b2 c3 − b3 c2 ) − a2 (b1 c3 − b3 c1 ) + a3 (b1 c2 − b2 c1 ).

We shall presently give formulas for the determinant of an

n × n matrix, that is, of a matrix of size n, where n ∈ N, and
we shall explore their use in matrix theory.

Saurav Bhaumik, IIT Bombay MA 110: Lecture 07

  
a11 · · · a1k ··· a1n
 .. .. .. 
 . . . 
 
 aj1 · · · ···
ajk ajn  n×n
Let n ∈ N and A :=  ∈R .
 .. .. .. 
 . . . 
an1 · · · ank ··· ann
The determinant of A is a real number defined inductively as
follows. For n := 1, define det A := a11 . Let n ≥ 2, and
suppose we have defined the determinant of any
(n − 1)×(n − 1) matrix. For j, k = 1, . . . , n, let Ajk denote
the submatrix of A obtained by deleting the jth row and the
kth column of A, and let Mjk := det Ajk , called the (j, k)th
minor of A. Define

det A := a11 M11 −a12 M12 +· · ·+(−1)1+k a1k M1k +· · ·+(−1)1+n a1n M1n .

This is also known as the expansion for the determinant of

A in terms of the first row of A.
Saurav Bhaumik, IIT Bombay MA 110: Lecture 07
An immediate consequence of our definition is the following.
Proposition
If an n × n matrix A is lower triangular, then the determinant
of A is the product of its diagonal entries.

Proof. We prove it by induction on n. For n = 1, it is the

definition. Suppose we know the proposition for all
(n − 1) × (n − 1) matrices. Now det A = a11 M11 since
a12 = · · · = a1n = 0, etc. Now A1,1 is lower triangular and
thus by the induction hypothesis, M1,1 is the product of the
diagonal entries of A other than a11 .

Saurav Bhaumik, IIT Bombay MA 110: Lecture 07

Next, it can be proved by induction on the size n of a matrix
that det A is equal to the following expansions in terms of the
jth row of A, and also in terms of the kth column of A:
n
X
det A = (−1)j+k ajk Mjk for each j ∈ {1, . . . , n}
k=1
n
X
det A = (−1)j+k ajk Mjk for each k ∈ {1, . . . , n}
j=1

(For a proof, see Kreyszig, Appendix 4, page A81.)

Saurav Bhaumik, IIT Bombay MA 110: Lecture 07

Proposition
Let A be a square matrix. Then det AT = det A.

Proof. This is obvious if n = 1. Let now n ≥ 2, and assume

this property for all (n − 1)×(n − 1) matrices. Note that
(AT )jk = (Akj )T for all j, k = 1, . . . , n, that is, the submatrix
obtained by deleting the jth row and the kth column of AT is
the same as the transpose of the submatrix obtained by
deleting the kth row and the jth column of A. (For example,
   
a11 a12 a13 a11 a21 a31
A := a21 a22 a23  =⇒ AT =  a12 a22 a32  ,
a31 a32 a33 a13 a23 a33
   T
T a21 a31 a21 a23
so that (A )21 = = = (A12 )T .)
a23 a33 a31 a33

Saurav Bhaumik, IIT Bombay MA 110: Lecture 07

 ′ ′
Let A := [ajk ] and AT := [ajk ]. Then ajk = akj and
′ T T
Mjk := det(A )jk = det(Akj ) = det Akj = Mkj by the
inductive hypothesis for j, k = 1, . . . , n. Expanding det AT in
terms of its first row, and det A in terms of its first column,
′ ′ ′ ′ ′ ′
det AT = a11 M11 − a12 M12 + · · · + (−1)1+n a1n M1n
= a11 M11 − a21 M21 + · · · + (−1)n+1 an1 Mn1
= det A.

Corollary
If A is upper triangular, then the determinant of A is the
product of its diagonal entries.
Proof. Let A be upper triangular. Then AT is lower triangular
and has the same diagonal entries as those of A.
 
Let us write A := c1 · · · ck · · · cn in terms of its n
columns.
Saurav Bhaumik, IIT Bombay MA 110: Lecture 07
Crucial Properties of the determinant function A 7−→ det A
from Rn×n to R:
1. Let k ∈ {1, . . . , n} and ck = α c′k + β c′′k , where α, β ∈ R.
Then det A = det c1 · · · α c′k + β c′′k · · · cn is equal to
 

α det c1 · · · c′k · · · cn + β det c1 · · · c′′k · · · cn .



This is proved by expanding det A in terms of its kth column.

In particular,  
det(αA) = det αc1 · · · αck · · · αcn = αn det A.
2. Let n ≥ 2. If ck = cℓ for some k ̸= ℓ, then det A = 0, that
is, if 2 columns of a matrix are identical, then its determinant
is equal to 0. This is clear for n = 2, and for n ≥ 3, this is
proved by expanding det A in terms of a column cp of A,
where p ̸= k and p ̸= ℓ, and then using induction on n.
3. det I = 1 , that is, the determinant of the identity matrix
is equal to 1. This is obvious.
Saurav Bhaumik, IIT Bombay MA 110: Lecture 07
Proposition
Let A be a square matrix.
(i) If two columns of A are interchanged, then det A gets
multiplied by −1.
(ii) Addition of a multiple of a column to another column of A
does not alter det A.
(iii) Multiplication of a column of A by a scalar α results in
the multiplication of det A by α.
 
Proof: Let A := c1 · · · ck · · · cℓ · · · cn , where k ̸= ℓ.
 
(i) Define α := det c1 · · · (ck + cℓ ) · · · (ck + cℓ ) · · · cn .
Then α = 0 since the matrix has two identical columns.
On the other
 hand, α = β + γ, where 
β := det  c1 · · · (ck + cℓ ) · · · ck · · · cn and
γ := det c1 · · · (ck + cℓ ) · · · cℓ · · · cn .
Saurav Bhaumik, IIT Bombay MA 110: Lecture 07
In turn, β := β1 + β2 and γ = γ1 + γ2 , where
 
β1 = det c1 · · · ck · · · ck · · · cn ,
 
β2 = det c1 · · · cℓ · · · ck · · · cn ,
 
γ1 = det c1 · · · ck · · · cℓ · · · cn ,
 
γ2 = det c1 · · · cℓ · · · cℓ · · · cn .

But β1 = 0 = γ2 since two columns are identical. Since

0 =α = β + γ = β2 + γ1 , we see that γ1 = −β2 , that is,
det c1 · · · ck · · · cℓ · · · cn is equal to
− det c1 · · · cℓ · · · ck · · · cn , as desired.
(ii) Suppose α times the ℓth column of A is added to the kth
column of A. Then det c1 · · · (ck + α cℓ ) · · · cℓ · · · cn is
equal to det A + α det c1 · · · cℓ · · · cℓ · · · cn = det A.
(iii)Suppose the kth column
 of A is multiplied by α. Then
det c1 · · · α ck · · · cn = α det c1 · · · ck · · · cn = α det A.
Saurav Bhaumik, IIT Bombay MA 110: Lecture 07