MA 110:

Lecture 08

Saurav Bhaumik,
Department of Mathematics
IIT Bombay

Spring 2025

Saurav Bhaumik, IIT Bombay MA 110: Lecture 08

Corollary
Let A be a square matrix.
(i) If two rows of A are interchanged, then det A gets
multiplied by −1.
(ii) Addition of a multiple of a row to another row of A does
not alter det A.
(iii) Multiplication of a row of A by a scalar α results in the
multiplication of det A by α.

Proof. Since the columns of AT are the rows of A, and since

det A = det AT , these results follow from the previous
proposition.
The above corollary can be used to find det A as follows.
Transform A to A′ by EROs of type I and type II, where A′ is
in REF, keeping track of the number of row interchanges.
Saurav Bhaumik, IIT Bombay MA 110: Lecture 08
Now A′ is an upper triangular matrix. Let p be the number of
′ ′
row interchanges, and let a11 , . . . , ann be the diagonal entries
′ ′ ′ ′
of A . Then det A = (−1) det A = (−1)p a11
p
· · · ann .
Corollary
A square matrix A is invertible if and only if det A ̸= 0.

Proof. Note that A is invertible if and only if A′ is. Now as A′

is a square matrix in REF, it is invertible if and only if there
are no zero rows. If there are no zero rows then each column
is pivotal with pivots given by precisely the diagonal entries.
Thus the formula above shows that det A = det A′ ̸= 0.
Conversely, if the k-th row is zero in A′ , then the diagonal
term akk = 0. Then the same formula above shows that
det A = det A′ = 0.

Saurav Bhaumik, IIT Bombay MA 110: Lecture 08

Example
   
0 2 0 −1 1 2 1 −1
1 2 1 −1 R1 ←→R2 0
  2 0 −1
Let A := 
0 0 −−−−−→  
3 2 0 0 3 2
1 −2 1 −2 1 −2 1 −2
   
1 2 1 −1 1 2 1 −1
R4 −R1 
0 2 0 −1 R4 +2R2 
 0 2 0 −1
 = A′ .
−− −→  −−−−→ 
0 0 3 2  0 0 3 2
0 −4 0 −1 0 0 0 −3
Since there is only one row interchange while transforming A
to A′ , since A′ is in REF, and since
det A′ = 1 · 2 · 3 · (−3) = −18, we see that
det A = (−1)(−18) = 18.

Saurav Bhaumik, IIT Bombay MA 110: Lecture 08

Remark We have given several criteria for the invertibility of
an n×n matrix A. We list them in the order we proved them.
(i) nullity(A) = 0.
(ii) There is a matrix B such that either BA = I or AB = I.
(iii) The RCF of A is I.
(iv) rank A = n.
(v) det A ̸= 0.

Saurav Bhaumik, IIT Bombay MA 110: Lecture 08

Determinant and Rank
We now relate the rank of a matrix with determinants of its
submatrices.
Lemma
Let A be an m × n matrix, and r ∈ N. Then

rank A ≥ r ⇐⇒ ∃ an r × r submatrix B of A with det B ̸= 0

Proof. Suppose rank A ≥ r . Since rank A equals the column

rank of A, there are r linearly independent columns of A. Let
C denote the m × r submatrix of A consisting of these r
columns. Then the column rank of C is r , and so the row rank
of C is also r . Hence there are r linearly independent rows of
C. Let B denote the r × r submatrix of C consisting of these
r rows. These r rows of B are linearly independent, and so
rank B = r . Hence B is invertible, and so det B ̸= 0.
Saurav Bhaumik, IIT Bombay MA 110: Lecture 08
Conversely, suppose B is an r × r submatrix of A such that
det B ̸= 0. Then B is invertible, and so rank B = r . Hence the
r rows of B, and consequently, the corresponding r rows of A
are linearly independent. Hence rank A ≥ r .

Corollary (Determinantal Characterization of Rank)

Let A be an m × n matrix, and r ∈ N. Then r = rank A if and
only if the following two conditions are satisfied.
(i) there is an r × r submatrix B of A such that det B ̸= 0,
(ii) det C = 0 for every (r + 1)×(r + 1) submatrix C of A.

Proof. This is an immediate consequence of the above lemma.

We remark that although the above result is of theoretical
interest, it does not give a practically useful method for finding
the rank of a matrix A. On the other hand, transformation of
A to a Row Echelon Form quickly reveals its rank.
Saurav Bhaumik, IIT Bombay MA 110: Lecture 08
Example  
3 0 2 2  
3 0
Let A := −6 42 24 54 . Since det
  ̸= 0
−6 42
21 −21 0 −15
and since the determinants of all 3 × 3 submatrices of A are
equal to 0, we see that rank A = 2.
This also follows by noting that A can be transformed by
EROs to  
3 0 2 2
A′ = 0 42 28 58 ,
0 0 0 0
which is in REF, and by noting that rank A is equal to the row
rank of A′ , which is 2.

Saurav Bhaumik, IIT Bombay MA 110: Lecture 08

We now consider another classical method of finding solutions
of Ax = b, where A is invertible.
Proposition (Cramer’s Rule)
 
Let A := c1 · · · ck · · · cn ∈ Rn×n  be invertible, and let

n×1
b ∈ R . For k = 1, . . . , n, let Bk := c1 · · · b · · · cn
be the matrix obtained by replacing the kth column ck of A by
the right side b of Ax = b. Then the unique solution
 T
x := x1 · · · xn of the linear system Ax = b is given by
det Bk
xk := for k = 1, . . . , n.
det A

Proof. If Ax = b, then b = x1 c1 + · · · + xk ck + · · · + xn cn . By
the first two crucial properties of the determinant function, 
det Bk = det c1 · · · x1 c1 + · · · + xk ck + · · · + xn cn · · · cn
= xk det A for k = 1, . . . , n. Since A is invertible, det A ̸= 0.
Hence the result.
Saurav Bhaumik, IIT Bombay MA 110: Lecture 08
    
3 −2 1 13
Let A := −2 1 4 , b :=  11 . Then det A = −60.
1 4 −5 −31
Also,  
13 −2 1
det B1 = det  11 1 4  = −60,
−31 4 −5
3 13 1
det B2 = det −2 11 4  = 180,
 1 −31 −5
3 −2 13
det B3 = det −2 1 11  = −240.
1 4 −31
Hence the unique solution of the linear system Ax = b is given
 T
by x1 := 1, x2 = −3, x3 = 4 , that is, x = 1 −3 4 .
Note: Cramer’s Rule is rarely used for solving linear systems;
the preferred method is the GEM. But Cramer’s Rule is of
theoretical interest, especially in solutions of differential eqns.
Saurav Bhaumik, IIT Bombay MA 110: Lecture 08
Formula for the Inverse of a Matrix
Let A := [ajk ] ∈ Rn×n with n ≥ 2. Recall that Ajk denotes the
submatrix of A obtained by deleting the jth row and the kth
column of A, and Mjk := det Ajk , the (j, k)th minor of A, for
j, k = 1, . . . , n. We define Cjk := (−1)j+k Mjk , j, k = 1, . . . , n.
It is called the cofactor of the entry ajk . Then the expansion
of det A in terms of the kth column is given by
Xn
det A = aℓk Cℓk , where k ∈ {1, . . . , n}.
ℓ=1

Define C := [Cjk ] ∈ Rn×n. It is called the cofactor matrix of A.

Theorem
Let A be a square matrix. Then CT A = (det A)I = ACT .
In particular, if det A ̸= 0, then A is invertible and

A−1 = CT / det A.
Saurav Bhaumik, IIT Bombay MA 110: Lecture 08
Proof. Let D := CT A = [djk ] say. By the definition
P of matrix
multiplication, the (j, k)th entry of D is djk = nℓ=1 Cℓj aℓk .
If j = k, then dkk = nℓ=1 Cℓk aℓk = det A, being the expansion
P
in terms of its kth column of A.
 
Let now j ̸= k. Write A := c1 · · · ck · · · cj · · · cn in
terms of its columns, and let B denote the matrix obtained by
replacingthe jth column cj by the kth column ck of A, that
is, B := c1 · · · ck · · · ck · · · cn = [bjk ], say. Then
det B = 0 since two columns are P identical. Expanding
Pn det B in
n
terms of its P
jth column, det B = ℓ=1 bℓj Cℓj = ℓ=1 aℓk Cℓj .
Thus djk = nℓ=1 Cℓj aℓk = det B = 0 if j ̸= k.
This shows that CT A = (det A)I. Similarly, we can prove
ACT = (det A)I, and so CT A = (det A)I = ACT .
In case det A ̸= 0, we see that A is invertible, and
A−1 = CT / det A.

Saurav Bhaumik, IIT Bombay MA 110: Lecture 08

Multiplicativity of Determinant Function
Proposition
Let A, B be n×n matrices. Then det(AB) = (det A)(det B).

Proof. Suppose first A is not invertible. Then (det A) = 0.

Also, AB is not invertible; otherwise there would be C such
that (AB)C = I, that is, A(BC) = I, which is impossible since
A is not invertible. Hence det(AB) = 0 = (det A)(det B).
Next, suppose A is invertible. Then we can transform A to a
diagonal matrix A′ (having nonzero diagonal elements) by
elementary row transformations of type I and type II. Then
det A′ = det A if an even number of row interchanges are
involved, and det A′ = − det A otherwise.
We observe that the same elementary row operations
transform AB to A′ B.
Saurav Bhaumik, IIT Bombay MA 110: Lecture 08
To see this, we can use Q. 2.3 in Tut Sheet 2: Making an
elementary row operation is equivalent to multiplying on the
left by the corresponding elementary matrix! And of course
E(AB) = (EA)B where E is any elementary matrix.
Hence det A′ B = det AB if an even number of row
interchanges are involved, and det A′ B = − det AB otherwise.
Thus it is enough to show that
det(AB) = (det A)(det B) when A is a diagonal matrix.
But this is easily seen because
     
α1 0 ··· 0 b1 α1 b 1
0 α2 · · · 0  b2   α2 b 2 
A :=  .. and B := =⇒ AB :=  ..  ,
     
.. .. ..   .. 
. . . .   .   . 
0 0 · · · αn bn αn b n
where b1 , . . . , bn denote the rows of B. Hence
det(AB) = α1 α2 · · · αn det B = (det A)(det B).
Saurav Bhaumik, IIT Bombay MA 110: Lecture 08
Corollary
1
If A is invertible, then det A−1 = .
det A
Proof. AA−1 = I =⇒ (det A)(det A−1 ) = det I = 1.
Example  
13 0 0
1 1
Let A :=  11 1 0 . Then det A−1 = =− .
det A 65
−31 22 −5

Saurav Bhaumik, IIT Bombay MA 110: Lecture 08

Linear Transformations
Just as we can define a continuous function from a subset of
Rn to Rm , we now define a ‘linear’ function from a subspace
of Rn×1 to Rm×1 .
Let V be a subspace of Rn×1 , and let W be a subspace of
Rm×1 . A linear transformation or a linear map from V to
W is a function T : V → W which ‘preserves’ the operations
of addition and scalar multiplication, that is, for all x, y ∈ V
and α ∈ R,
T (x + y) = T (x) + T (y) and T (α x) = α T (x).
It follows that if T : V → W is linear, then T (0) = 0, and
T ‘preserves’ linear combinations of vectors in V , that is,
T (α1 x1 + · · · + αk xk ) = α1 T (x1 ) + · · · + αk T (xk )
for all x1 , . . . , xk ∈ V and α1 , . . . , αk ∈ R.
Saurav Bhaumik, IIT Bombay MA 110: Lecture 08
Model Example
Let V := Rn×1 , W := Rm×1 and A be an m×n matrix, that is,
A ∈ Rm×n . Define TA : Rn×1 → Rm×1 by

TA (x) = A x for x ∈ V .

The properties of matrix multiplication show that TA is linear.

Conversely, suppose T : Rn×1 → Rm×1 is linear. We show that
 T
T = TA for some A ∈ Rm×n. Let x := x1 · · · xn ∈ Rn×1 .
Then x = x1 e1 + · · · + xn en , where e1 , . . . , en are the basic
column vectors in Rn×1 . Since T is linear, we obtain

T (x) = x1 T (e1 ) + · · · + xn T (en ).

 
Define ck := T (ek ) for k = 1, . . . , n, and A := c1 · · · cn .
Then T (x) = x1 c1 + · · · + xn cn = A x for all x ∈ Rn×1 . Thus
A ∈ Rm×n and T = TA . (Note: kth column of A is T (ek ).)
Saurav Bhaumik, IIT Bombay MA 110: Lecture 08
Thus every linear transformation T : Rn×1 → Rm×1 is given by
      
x1 a11 x1 + · · · + a1n xn x1
 ..  .
.. . n×1
T  .  :=   for  ..  ∈ R ,
  
xn am1 x1 + · · · + amn xn xn

where a11 , . . . , a1n , . . . , am1 , . . . , amn ∈ R.

Similarly, one can define a linear map T : R1×n → R1×m , and
find that for x1 · · · xn ∈ R1×n ,
 
 
T x1 · · · xn := a11 x1 +· · ·+ a1n xn · · · am1 x1 +· · ·+ amn xn .

Remark: Let D be an open subset of R1×2 , [x0 , y0 ] ∈ D, and

let a function f : D → R be differentiable at [x0 , y0 ]. Then the
total derivative of f at [x0 , y0 ] is a linear map (which depends
on f ) given by T ([x, y ]) = αx + βy for [x, y ] ∈ R1×2 , where
α := fx (x0 , y0 ) and β := fy (x0 , y0 ).
Saurav Bhaumik, IIT Bombay MA 110: Lecture 08
Let A, B ∈ Rm×n and α, β ∈ R. Then αA + βB ∈ Rm×n and
TαA+βB (x) = (αA + βB)x = αTA (x) + βTB (x)
for x ∈ Rn×1 . We write this as follows:
TαA+βB = αTA + βTB .

Next, let A ∈ Rm×n and B ∈ Rn×p . Then AB ∈ Rm×p , and

TAB (x) = (AB)x = A(Bx) = TA (Bx) = TA (TB (x)) = TA ◦TB (x)
for x ∈ Rp×1 by the associativity of matrix multiplication. Thus
TAB = TA ◦ TB .
This says that the linear map associated with the product AB
of matrices A and B is the composition of the linear maps
associated with A and associated with B in the same order.
This partially justifies the definition of matrix multiplication.
Saurav Bhaumik, IIT Bombay MA 110: Lecture 08
Examples
Let A ∈ R2×2 . Then TA : R2×1 → R2×1 .
 
2 0  T  T
(i) Let A := . Then TA : x1 x2 7−→ 2x1 2x2 .
0 2
TA stretches each vector by a factor of 2.
 
0 1  T  T
(ii) Let A := . Then TA : x1 x2 7−→ x2 x1 .
1 0
TA is the reflection in the line x1 = x2 .
 
−1 0  T  T
(iii) Let A := .Then TA : x1 x2 7→ −x1 −x2 .
0 −1
TA is the reflection in the origin.
 
cos θ − sin θ
(iv) Let A := , where θ ∈ (−π, π]. Then
sin θ cos θ
 T  T
TA : x1 x2 7−→ x1 cos θ − x2 sin θ x1 sin θ + x2 cos θ .
TA is the rotation through an angle θ.
These are geometric interpretations of matrices.
Saurav Bhaumik, IIT Bombay MA 110: Lecture 08