MATH 304

Linear Algebra
Lecture 8:
Inverse matrix (continued).
Elementary matrices.
Transpose of a matrix.
Inverse matrix

Definition. Let A be an n×n matrix. The inverse

of A is an n×n matrix, denoted A−1 , such that
AA−1 = A−1 A = I .
If A−1 exists then the matrix A is called invertible.
Otherwise A is called singular.
Inverting diagonal matrices

Theorem A diagonal matrix D = diag(d1 , . . . , dn )

is invertible if and only if all diagonal entries are
nonzero: di 6= 0 for 1 ≤ i ≤ n.
If D is invertible then D −1 = diag(d1−1 , . . . , dn−1 ).

 −1  
d1 0 . . . 0 d1−1 0 ... 0
0 d ... 0   0 d −1 ... 0 
2 2
=
   
 .. .. . . .
.  .. .. . . . ... 
. . . .  . .


−1
0 0 . . . dn 0 0 . . . dn
Inverting 2-by-2 matrices

Definition.
 The determinant of a 2×2 matrix
a b
A= is det A = ad − bc.
c d
 
a b
Theorem A matrix A = is invertible if
c d
and only if det A 6= 0.
If det A 6= 0 then
 −1  
a b 1 d −b
= .
c d ad − bc −c a
Fundamental results on inverse matrices

Theorem 1 Given a square matrix A, the following are

equivalent:
(i) A is invertible;
(ii) x = 0 is the only solution of the matrix equation Ax = 0;
(iii) the row echelon form of A has no zero rows;
(iv) the reduced row echelon form of A is the identity matrix.

Theorem 2 Suppose that a sequence of elementary row

operations converts a matrix A into the identity matrix.
Then the same sequence of operations converts the identity
matrix into the inverse matrix A−1 .

Theorem 3 For any n×n matrices A and B,

BA = I ⇐⇒ AB = I .
Row echelon form of a square matrix:
   
∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗
   

 ∗ ∗ ∗ ∗ ∗ 


 ∗ ∗ ∗ ∗ ∗ 

∗ ∗ ∗ ∗  ∗ ∗ ∗ ∗ 
   
 
   
∗ ∗ ∗  ∗ ∗ 
   
 
   

 ∗ ∗ 


 ∗ 

   

 ∗ 






invertible case noninvertible case

  
3 −2 0
Example. A =  1 0 1.
−2 3 0
To check whether A is invertible, we convert it to
row echelon form.
Interchange the 1st row with the 2nd row:
 
1 0 1
 3 −2 0
−2 3 0
Add
 −3 times the
 1st row to the 2nd row:
1 0 1
 0 −2 −3
−2 3 0
Add 2 times the 1st row to the 3rd row:
 
1 0 1
0 −2 −3
0 3 2

Multiply the 2nd row by −1/2:

 
1 0 1
0 1 1.5
0 3 2
Add −3 times the 2nd row to the 3rd row:
 
1 0 1
0 1 1.5
0 0 −2.5
Multiply the 3rd row by −2/5:
 
1 0 1
 0 1 1.5
0 0 1
We already know that the matrix A is invertible.
Let’s proceed towards reduced row echelon form.
Add −3/2 times the 3rd row to the 2nd row:
 
1 0 1
0 1 0
0 0 1
Add
 −1 times
 the 3rd row to the 1st row:
1 0 0
0 1 0
0 0 1
To obtain A−1 , we need to apply the following
sequence of elementary row operations to the
identity matrix:
• interchange the 1st row with the 2nd row,
• add −3 times the 1st row to the 2nd row,
• add 2 times the 1st row to the 3rd row,
• multiply the 2nd row by −1/2,
• add −3 times the 2nd row to the 3rd row,
• multiply the 3rd row by −2/5,
• add −3/2 times the 3rd row to the 2nd row,
• add −1 times the 3rd row to the 1st row.
A convenient way to compute the inverse matrix
A−1 is to merge the matrices A and I into one 3×6
matrix (A | I ), and apply elementary row operations
to this new matrix.
   
3 −2 0 1 0 0
A =  1 0 1 , I = 0 1 0
−2 3 0 0 0 1
 
3 −2 0 1 0 0
(A | I ) =  1 0 1 0 1 0
−2 3 0 0 0 1
 
3 −2 0 1 0 0
 1 0 1 0 1 0
−2 3 0 0 0 1
Interchange the 1st row with the 2nd row:
 
1 0 1 0 1 0
 3 −2 0 1 0 0
−2 3 0 0 0 1
Add −3 times the 1st row to the 2nd row:
 
1 0 1 0 1 0
 0 −2 −3 1 −3 0
−2 3 0 0 0 1
Add 2 times the 1st row to the 3rd row:
 
1 0 1 0 1 0
0 −2 −3 1 −3 0
0 3 2 0 2 1

Multiply the 2nd row by −1/2:

 
1 0 1 0 1 0
0 1 1.5 −0.5 1.5 0
0 3 2 0 2 1
Add −3 times the 2nd row to the 3rd row:
 
1 0 1 0 1 0
0 1 1.5 −0.5 1.5 0
0 0 −2.5 1.5 −2.5 1
Multiply the 3rd row by −2/5:
 
1 0 1 0 1 0
0 1 1.5 −0.5 1.5 0
0 0 1 −0.6 1 −0.4

Add −3/2 times the 3rd row to the 2nd row:

 
1 0 1 0 1 0
0 1 0 0.4 0 0.6
0 0 1 −0.6 1 −0.4
Add −1 times the 3rd row to the 1st row:
 
1 0 0 0.6 0 0.4
0 1 0 0.4 0 0.6
0 0 1 −0.6 1 −0.4
 −1 3 2
 

3 −2 0 5 0 5
 2 3
Thus  1 0 1 =  5 0 5 .
−2 3 0 − 35 1 − 25
That is,
  3 0 2
  
3 −2 0 5 5 1 0 0
 1 0 1  2 
0 3
= 0 1 0,
5 5
−2 3 0 − 35 1 − 25 0 0 1
 3 2

5 0 5 3 −2 0
 
1 0 0

 2 3 
 5 0 5 1 0 1 = 0 1 0.
− 35 1 − 25 −2 3 0 0 0 1
Why does it work?
    
1 0 0 a1 a2 a3 a1 a2 a3
0 2 0 b1 b2 b3  = 2b1 2b2 2b3 ,
0 0 1 c1 c2 c3 c1 c2 c3
    
1 0 0 a1 a2 a3 a1 a2 a3
3 1 0b1 b2 b3 =b1 +3a1 b2 +3a2 b3 +3a3 ,
0 0 1 c1 c2 c3 c1 c2 c3
    
1 0 0 a1 a2 a3 a1 a2 a3
 0 0 1   b 1 b 2 b 3  =  c1 c2 c 3 .
0 1 0 c1 c2 c3 b1 b2 b3
Proposition Any elementary row operation can be
simulated as left multiplication by a certain matrix.
Elementary matrices

 
1
 ... O 
1
 
 
E = r  row #i
 
 1 
...
 
 O 
1

To obtain the matrix EA from A, multiply the ith

row by r . To obtain the matrix AE from A, multiply
the ith column by r .
Elementary matrices

1
 
 ... ... O 
0 ··· 1 row #i
 
E =  ... .. . . .

.
 

0 ··· r ··· 1  row #j
 ... .. .. . . .
 
. . 
0 ··· 0 ··· 0 ··· 1

To obtain the matrix EA from A, add r times the ith

row to the jth row. To obtain the matrix AE from
A, add r times the jth column to the ith column.
Elementary matrices

1
 
O
 ... 
0 ··· 1 row #i
 
.. . . . ..
 
E = . .
 

 1 ··· 0  row #j
...
 
 
O 1

To obtain the matrix EA from A, interchange the

ith row with the jth row. To obtain AE from A,
interchange the ith column with the jth column.
Why does it work?

Assume that a square matrix A can be converted to

the identity matrix by a sequence of elementary row
operations. Then
Ek Ek−1 . . . E2 E1 A = I ,
where E1 , E2 , . . . , Ek are elementary matrices
corresponding to those operations.

Applying the same sequence of operations to the

identity matrix, we obtain the matrix
B = Ek Ek−1 . . . E2 E1 I = Ek Ek−1 . . . E2 E1 .
Thus BA = I , which implies that B = A−1 .
Transpose of a matrix

Definition. Given a matrix A, the transpose of A,

denoted AT , is the matrix whose rows are columns
of A (and whose columns are rows of A). That is,
if A = (aij ) then AT = (bij ), where bij = aji .
 
 T 1 4
1 2 3
Examples. =  2 5 ,
4 5 6
3 6
 T
7  T  
8 = (7, 8, 9), 4 7 4 7
= .
7 0 7 0
9
Properties of transposes:
• (AT )T = A
• (A + B)T = AT + B T
• (rA)T = rAT
• (AB)T = B T AT
• (A1 A2 . . . Ak )T = ATk . . . AT2 AT1
• (A−1 )T = (AT )−1
Definition. A square matrix A is said to be
symmetric if AT = A.
For example, any diagonal matrix is symmetric.

Proposition For any square matrix A the matrices

B = AAT and C = A + AT are symmetric.
Proof:
B T = (AAT )T = (AT )T AT = AAT = B,
C T = (A + AT )T = AT + (AT )T = AT + A = C .