Equivalent statements to 'A is invertible'.

One way to determine if A1 exists and how to

find it if it does.
Proof of the '50% Theorem'.
Definition of the determinant of a square matrix.
Cofactor expansion (along any row or column).

Three theorems about determinants

(transposes, triangular matrices, matrices with
identical rows or columns).
Determinants (cont’d)
 a b c 
  We already know that
A d e f
 
 g h i  det(A)  aei  bfg  cdh  ceg  afh  bdi
 
kR3 Recall that E1 A  B1 where
1 0 0
 a b c
E1   0 1 0 
B1   d e f  
0 0 k 

   
 kg kh ki 
 
det(B1)  aeki  bfkg  cdkh  cekg  afkh  bdki
 k (aei  bfg  cdh  ceg  afh  bdi)  kdet(A)
 a b c  kR3  a b c
Ad e f   d e f B
    1
g h i   kg kh ki 
   
1 0 0
E1 A  B1 where E1   0 1 0  det(E1)  k
 
0 0 k 
  det(B1)  kdet( A)
So it seems like elementary
 det( E1)det(A)
row operations of this type
changes the determinant of A by a factor of k and
we have 'determinant of product equals to product of
determinants' (to some extent).
 a b c 
  We already know that
A d e f
 
 g h i  det(A)  aei  bfg  cdh  ceg  afh  bdi
 
R1  R3 Recall that E2 A  B2 where
0 0 1
g h i 
E2   0 1 0 
B2   d e f  
1 0 0

   
a b c 
 
det(B2 )  gec  hfa  idb  iea  gfb  hdc
 (iea  gfb  hdc  gec  hfa  idb)  det(A)
 a b c  R R g h i 

A d e f  1 3 
d e f   B2
   
g h i  a b c 
   
0 0 1
E2 A  B2 where E2   0 1 0  det(E2 )  1
 
1 0 0
  det(B2 )  det( A)
So it seems like elementary
 det( E2 )det(A)
row operations of this type
changes the determinant of A by a factor of  1 and
we have 'determinant of product equals to product of
determinants' (to some extent).
 a b c 
  We already know that
A d e f
 
 g h i  det(A)  aei  bfg  cdh  ceg  afh  bdi
 
R2  kR3 Recall that E3 A  B3 where
1 0 0
 a c 
E3   0 1 k 
b
B3   d  kg e  kh f  ki  
0 0 1

   
 g i 
 h

det(B3 )  ai(e  kh)  bg ( f  ki)  ch(d  kg )

cg (e  kh)  ah( f  ki)  bi(d  kg )  det(A)
  a b c  R  kR  a b c 

A d e f  2 3 
d  kg e  kh f  ki   B3
   
g h i   g 
   h i 
1 0 0
E3 A  B3 where E3   0 1 k  det(E3 )  1 (why?)
 
0 0 1
  det(B3 )  det( A)
So it seems like elementary
 det( E3 )det(A)
row operations of this type
does not change the determinant of A and
we have 'determinant of product equals to product of
determinants' (to some extent).
Let A be a square matrix. Then

B1 det(B1)  kdet( A)
kRi

Ri  R j
A B2 det(B2 )  det( A)

R j  kRi
B3 det(B3 )  det( A)

Furthermore, if E is an elementary matrix of the

same size as A, then det( EA)  det( E )det( A).
 We will prove the following:
R j  kRi
A B3 det(B3 )  det( A)
Let A* be the following
 a11 a12 ... ... a1n   a11 a12 ... ... a1n 
   
   
row i  ai1 ai 2 ... ... ain   ai1 ai 2 ... ... ain 
A  A*   
   
row j  a j1 a j 2 ... ... a jn   ai1 ai 2 ... ... ain 
   
   
a an 2 ... ... ann  a an 2 ... ... ann 
 n1  n1
 Observations:
1) A and A* differ in only one row (row j).
2) det(A* )  0 since A* has two identical rows.

 a11 a12 ... ... a1n   a11 a12 ... ... a1n 
   
   
row i  ai1 ai 2 ... ... ain   ai1 ai 2 ... ... ain 
A  A*   
   
row j  a j1 a j 2 ... ... a jn   ai1 ai 2 ... ... ain 
   
   
a an 2 ... ... ann  a an 2 ... ... ann 
 n1  n1
 For k  1, 2,..., n,
cofactors of A Ajk  Ajk *
Observations:
3) ( j ,1)-cofactor of A  ( j ,1)-cofactor of A*
cofactors of A*
( j , 2)-cofactor of A  ( j , 2)-cofactor of A*

 a11 a12 ... ... a1n   a11 a12 ... ... a1n 
   
   
row i  ai1 ai 2 ... ... ain   ai1 ai 2 ... ... ain 
A  A*   
   
row j  a j1 a j 2 ... ... a jn   ai1 ai 2 ... ... ain 
   
   
a an 2 ... ... ann  a an 2 ... ... ann 
 n1  n1
 For k  1, 2,..., n,
cofactors of A Ajk  Ajk *
Observations:
3) ( j ,1)-cofactor of A  ( j ,1)-cofactor of A*
cofactors of A*
( j , 2)-cofactor of A  ( j , 2)-cofactor of A*
cofactor expansion along
 a11 a12 ... ... a1n 
row j of A* :  
 
det( A* )  ai1 Aj1*  ai 2 Aj 2*   ai1 ai 2 ... ... ain 
...  ain Ajn* A*   
 
 ai1 Aj1  ai 2 Aj 2  ...  ain Ajn  ai1 ai 2 ... ... ain 
 
(by observation 3)  
a an 2 ... ... ann 
 0 (by observation 2)  n1
 ai1 Aj1  ai 2 Aj 2  ...  ain Ajn  0

 a11 a12 ... ... a1n   a11 a12 ... ... a1n 
   
   
 ai1 ai 2 ... ... ain   ai1 ai 2 ... ... ain 
   
   
 a j1 a j 2 ... ... a jn   a j1  kai1 ... ... ... a jn  kain 
   
   
a an 2 ... ... ann   a ann 
 n1  n1 an 2 ... ...
A B3
R j  kRi
 ai1 Aj1  ai 2 Aj 2  ...  ain Ajn  0

cofactor expansion  a11 a12 ... ... a1n 

along row j of B3 :  
 
det(B3 )  ai1 ai 2 ... ... ain 
 
 (a j1  kai1) Aj1  ...  
 (a jn  kain ) Ajn  a j1  kai1 ... ... ... a jn  kain 
 
 
 a ann 
 n1 an 2 ... ...
A B3
R j  kRi
 ai1 Aj1  ai 2 Aj 2  ...  ain Ajn  0
det(B3 )  (a j1  kai1) Aj1  ...  (a jn  kain ) Ajn
 a j1 Aj1  kai1 Aj1  ...  a jn Ajn  kain Ajn

 a j1 Aj1  a j 2 Aj 2  ...  a jn Ajn 

k (ai1 Aj1  ai 2 Aj 2  ...  ain Ajn )  0
 a j1 Aj1  a j 2 Aj 2  ...  a jn Ajn
 det( A) (cofactor expansion along row j of A)
A B3
R j  kRi
 suppose you want to find
wow, this is quite
the determinant of A...
something, but how
... you first find the
can we use such a
determinant of its row-
result?
echelon form.

A R
 Yes! R is a triangular matrix
oh I know, then we
whose determinant is easy
keep track of what
to evaluate... we can now
e.r.o. have been
'backtrack' to find the
performed on A...
determinant of A.

A R
Find the determinant of the following matrix using
elementary row operations.

1 3 1 1
2 5 2 2
 
1 3 8 9
1 2 
 3 2
Suppose A and B are row equivalent matrices as
shown below:
 5 0 8 1
R1  92 R2 R2  R3 4R2  0 2 1 0 
A  B
0 0 1 9 
0 0 0 1 
 3 

Find det( A).

det(B)  1 (1)  (4)  det( A)
5  (2) 1 13  4det( A)
Can you find A?
5
6  det( A)
Instead of performing elementary operations on rows
of a matrix, we can also perform these operations on
columns. These operations are called elementary
column operations.
Corresponding to the previous theorem, there is
another one stated in terms of elementary column
operations.
However, column operations are not discussed in this
module.
If A is a square matrix, then the following statements
are equivalent.
1) A is invertible.

2) Ax  0 has only the trivial solution.

3) The reduced row-echelon form of A is I .

4) A can be expressed as a product of

elementary matrices.
We can now add one more equivalent statement!
5) det( A)  0
A square matrix A is invertible if and only if
det( A)  0.

Recall that we have already proven this:

a b
A  is invertible if and only if
c d
ad  bc  det( A)  0.
Let A and B be two square matrices of order n
and c is a scalar.

1) det(cA)  c ndet( A)

Proof:
cR1 , cR2 ,..., cRn
A cA

Each cRi changes the determinant by a factor

of c , so det(cA)  c ndet( A).
Let A and B be two square matrices of order n
and c is a scalar.
2) det( AB)  det( A)det(B)

Remark:
This results generalizes one that we had previously:

det( EA)  det( E)det( A)

where E is an elementary matrix of the same

order as A.
2) det( AB)  det( A)det(B)
Proof:
If A is singular, we already know that AB is singular.
In this case, det( A)  0, det( AB)  0, so

0  det(AB)  det( A)det(B)  0

Next consider the case when A is invertible.

2) det( AB)  det( A)det(B)
Proof:
Next consider the case when A is invertible.
Since A is invertible, it can be expressed as a product
of elementary matrices.

A  Ek Ek1... E2 E1
So
AB  Ek Ek1... E2 E1B
 det( AB)  det( Ek Ek1... E2 E1B)
2) det( AB)  det( A)det(B)
Proof:
We use the result det(EA)  det( E )det(A)
repeatedly on

 det( AB)  det( Ek Ek1... E2 E1B)

 det( Ek )det( Ek1... E2 E1B)

 det( Ek )det( Ek1)...det(E2 )det(E1)det(B)

2) det( AB)  det( A)det(B)
Proof:
We again use the result det(EA)  det( E )det(A)
repeatedly on
det(AB)  det( Ek )det(Ek1)...det(E2 )det(E1)det(B)
 det( Ek )det(Ek1)...det(E2 E1)det(B)

 det( Ek Ek1... E2 E1)det(B)

 det( A)det(B)
Let A and B be two square matrices of order n
and c is a scalar.
1 1
3) If A is invertible, then det( A )  .
det( A)
Proof:
A1 A  I  det( A1 A)  det( I )
1 by det(AB)
 det( A )det( A)  det( I )
 det( A)det(B)
 det( A1)det( A)  1
1 1
 det( A ) 
det( A)
  3 2 4 
Let A   4 3 1 . We can check that det(A)  34.
 
 0 2 4
 

det(4 A)  43 det( A)  64  34  2176.

1 0 1
If B   0 1 0  , det(B) 
 
1 0 2
 

det( AB)  det( A1) 

Let A be a square matrix of order n. The adjoint
of A is the square matrix of order n
T
 A11 A12 ... A1n   A11 A21 ... An1 
 A A ... A   A A ... A 
adj( A)   21 22 2n
   12 22 n2

   
 A A ... A   A A ... A 
 n1 n2 nn   1n 2n nn 

Recall that Aij is the (i , j)-cofactor of A.

Let A be a square matrix of order n. If A is invertible,
then
1 1
A  adj( A).
det( A)

Proof:
'candidate'
It suffices to show that

 1 
A adj(A)  I
 det( A) 
 Proof:
 1 
A adj(A)  I
 det( A) 

(i , j)-entry of Aadj(A)

 a11 a12 ... ... a1n   A11 ... Aj1 ... An1 
   A ... A ... A 
 (i , j)    12 j2 n2

 
-entry of  ai1 ai 2 ... ... ain   
 
   
a a   A ... A ... A 
 n1 n 2 ... ... ann   1n jn nn 

 ai1 Aj1  ai 2 Aj 2  ...  ain Ajn

Proof:

(i , j)-entry of Aadj(A)  ai1 Aj1  ai 2 Aj 2  ...  ain Ajn

When i  j , (i ,i)-entry of Aadj( A) is

ai1 Ai1  ai 2 Ai 2  ...  ain Ain

 det( A) (cofactor expansion along row i)

When i  j , (i , j)-entry of Aadj( A) is

 ai1 Aj1  ai 2 Aj 2  ...  ain Ajn

 0 (see slide #13)

Proof:
det( A) if i  j
(i , j)-entry of Aadj(A)  
0 if i  j

 det( A) 
 1
A

adj(A) 
1 
 det( A) 0 

 det( A)  
 det( A)


0 det( A) 
I
11
So we have shown that A  adj(A).
det( A)
 a b
Let A    . Find adj( A).
c d

A11  d A12  c A21  b A22  a

 A11 A12   d c 
T T

adj( A)     
 21
A A22   b a 
 d b 
 
  c a 

1 1  d b 
A 
ad  bc  c a 
  3 2 4 
Let A   4 3 1 . We have seen that det(A)  34.
 
 0 2 4
 
3 1 4 1 4 3
A11   10 A12    16 A13  8
2 4 0 4 0 2

2 4 3 4 3 2
A21    16 A22   12 A23   6
2 4 0 4 0 2

2 4 3 4 3 2
A31   14 A32    19 A33   1
3 1 4 1 4 3
  10 16 8   10 16 14 
T

adj(A)   16 12 6    16 12 19 

   
 14 19 1  8  
   6 1 

 10 16 14 
1
1
A  16 12 19 
34  

 8 6 1 
Suppose Ax  b is a linear system where A is an
invertible square matrix of order n. For i  1, 2,..., n,
let Ai be the square matrix of order n where the
ith column of A is replaced by b.
Cramer's
 x1 
x  Rule
Then the unique solution is x   2  where
 
x 
 n
det( Ai )
xi  , i  1, 2,..., n.
det( A)
Solve the following linear system by Cramer's Rule.
 x  y  z  1

2 x  y  z  4
 x  2 y  3z  7


1 1 1   1 1 1   1 1 1 
A   2 1 1  A1   4 1 1  A2   2 4 1 
     
 1 2 3   7 2 3   1 7 3 
     
 1 1 1
A3   2 1 4 
 
1 2 7 
 
det( A)  15 det( A1)  15 det( A2 )  0 det( A3 )  30

15 0 30
x  1 y 0 z  2
15 15 15

1 1 1   1 1 1   1 1 1 
A   2 1 1  A1   4 1 1  A2   2 4 1 
     
 1 2 3   7 2 3   1 7 3 
     
 1 1 1
A3   2 1 4 
 
1 2 7 
 
 Lecture 08:
Euclidean n-spaces
Linear Combinations and Linear Spans
(till Example 3.2.8)