You are on page 1of 5

INVERSE of a MATRIX

The term inverse you've met many times already. The number 1 is the inverse of 3 in the sense that 3 × 1
= 1 , and it
3 3
was used to solve a simple equation like 3x = 4 . On the other hand, a function g is said to be the inverse of a function f
when f(g(x)) = x and g(f(x)) = x whenever the functions make sense. For instance, g(x) = log(x) is the inverse of
f(x) = ex because log(ex) = x and elog x = x, properties you've used many times! But think about expressing these as
g ∘ f = I and f ∘ g = I where I is the Identity function, meaning I(x) = x for all x. We are now going to extend these
ideas to matrices and matrix transformations.

The IDENTITY matrix is the n × n matrix In defined by

⎡1 0 0 ⋯ 0 0⎤
⎢0 1 0 ⋯ 0 0⎥
In = ⎢

⎥.
⋮⎥
⎢⋮ ⎥
⎢0 0⎥
⋮ ⋮ ⋯ ⋮
⎣0 1⎦
0 0 ⋯ 1
0 0 ⋯ 0
AIn = A for any m × n matrix A, while In B = B for any n × p matrix B; in other words,
It's called the identity because
it does for matrices what the number 1 does for multiplication of numbers.

Definition: an n × n matrix A is said to be Invertible when there exists an n × n


matrix B such that AB = In and BA = In where In is the n × n identity matrix. If A
is invertible it is often said to be non-singular, while A is said to be singular when it
is not invertible.

Notice that that invertibility is defined only for square matrices because AB, BA cannot both be defined unless A, B are
square.

Inverse Property 1: (Uniqueness) if A is Inverse Property 2: (Double Inverse) if A is


−1 −1
invertible, it can have only one inverse, and this is invertible, then (A ) = A.
−1
denoted by A .

Inverse Property 3: (2/3 Rule) if any two of Inverse Property 4: (Matrix Equation) if A is an
A, B, and AB are invertible, then so is the third, and invertiblen × n matrix A, then the matrix equation
−1
morever, (AB) = B−1 A−1 . Ax = b has a solution x = A−1 b for each b in
Rn , and this solution is unique.
In the case of 2×2 matrices the inverse is easy to compute. Recall that the determinant det(A) of a 2 × 2 matrix A
is given by

]) = ad − bc .
a b
det(A) = ∆(A) = ∆([
c d

A=[ ] be a 2 × 2
a b
Theorem: let such that ∆(A) ≠ 0. Then A is Invertible and
c d
1
[ ].
d −b
A−1 =
∆(A) −c a

If ∆(A) = 0 , then A is not invertible.

To prove the theorem simply carry out the multiplications:

[ ][ ] = [ ] = ∆(A)I2 ,
a b d −b ad − bc 0
c d −c a 0 ad − bc

and similarly for multiplication in the opposite order. A direct calculuation also produces the inverse of some n×n
matrices:

diagonal matrices: if the diagonal entries d1 , d2 , … , dn are all non-zero, then

⎡ d1 0 ⎤ ⎡ 1/d1 0 ⎤
−1 0 ⋯ 0
0 ⋯ 0
⎢0 0 ⎥ ⎢

0 1/d2 ⋯ 0 0 ⎥

⎢ ⎥
d2 ⋯ 0
⎢ ⎥ ⎢
= ⎢ ⋮ ⎥

⎢ ⋮ ⋮ ⎥
⎥ ⎢ ⋮ ⎥⎥
⎢ ⎥
⋮ ⋯ ⋮ ⋮ ⋯ ⋮
⎢0 0 ⎥ ⎢ 0 0 ⎥
⎣0 d ⎦
0 ⋯ dn−1 0 ⋯ 1/dn−1
0 ⋯ 0 n
⎣ 0 0 ⋯ 0 1/dn ⎦

triangular matrices:

⎡1 0 0⎤ ⎡ 0⎤ ⎡1 p q⎤ ⎡ 1 −p pr − q ⎤
−1 −1
1 0
⎣c d 1⎦ ⎣ db − c −d 1⎦
= 0 , = 0 1 .
⎣0 0 1⎦ ⎣0 0 1 ⎦
b 1 0 −b 1 0 1 r −r

There is no correspondingly simple formula for n × n matrices in general, however, though there is an algorithm for
computing the inverse. It's based on reducing a certain matrix to its reduced row echelon form.

Theorem: if A an n × n matrix and [A In ] can be row reduced to [In B] for some B,


−1
then A is invertible and B = A . Otherwise, A is singular. In matrix terms

rref([A In ]) = [In A−1 ] .


Thus by row reduction,
Example 1: show that if

⎡0 1 2⎤ ⎡1 0 3 0 1 0⎤
⎣ 4 −3 1⎦
[A I3 ] ∼ 0 1 2 1 0 0
⎣ 4 −3 8⎦
A = 1 0 3 ,
8 0 0

then ⎡1 0 3 0 1 0⎤
⎣ 0 −3 1⎦
∼ 0 1 2 1 0 0
⎡ −9/2 7 −3/2 ⎤ −4 0 −4
A−1 =
⎡1 0 0⎤
−1 .
⎣ 3/2 −2 1/2 ⎦
−2 4 3 0 1
⎣0 0 1⎦
∼ 0 1 2 1 0 0 .
12 3 −4

⎡1 0 0 −9/2 7 −3/2 ⎤
Solution: we could check by direct calculation, of ∼ 0 1 = rref([A I3 ]).
⎣0 0 ⎦
0 −2 4 −1
course, but let's use the algorithm to compute 1 3/2 −2 1/2
rref([A I3 ]) by hand (or use Wolfram Alpha!). We
Consequently,
omit some of the steps!

⎡1 0 0 −9/2 7 −3/2 ⎤
⎡0 1 2 1 0 0⎤ −1
[I3 A ] = 0 1 −1 .
⎣0 0 1/2 ⎦
0 −2 4
⎣ 4 −3 1⎦
[A I3 ] = 1 0 3 0 1 0
8 0 0 1 3/2 −2

To understand why such an algorithm for computing A−1 works brings us to another fundamental idea in the uses of
linear algebra - that's the subject of the next lecture! But first we need to see what's the connection between row reduction
and matrix inverses. Recall in more specific terms than before:

Elementary Row Operations on a matrix: these operations produce new rows R′j
in the matrix from one or more old rows Ri in the matrix:

(Replacement) replace row Rj by the sum of itself and a multiple cRi of row Ri:

Eij (c) : cRi + Rj → R′j

(Interchange) interchange rows Ri and Rj:

Eij : Ri → R′j , Rj → R′i ,

(Scaling) multiply all entries in row Ri by a non-zero constant c:

Ei(c) : cRi → R′i .

But these are invertible as operators from new rows back to old rows:

Eij (c)−1 : (−c)R′i + R′j → Ri , Eij−1 : R′2 → R1 , R′1 → R2 , Ei(c)−1 : ( 1c )R′i → Ri .

Consequently,

Eij (c)−1 = Eij (−c), Eij−1 = Eji, Ei(c)−1 = Ei( 1c ) .


Thus elementary row operations will provide us with lots of important examples of invertible matrices.

An m × m matrix E said to be an Elementary Matrix when its product


EA with every m × n matrix A is an elementary row operation on A. We'll
use the same E-notation for the Elementary matrix corresponding to an
Elementary row operation.

The 3×3 Elementary matrices are typical of the general case - we've met may of them already!! For concreteness, set

⎡ 0 1 1 4 ⎤ ⎡1 0 0 1⎤
⎣ −2 −3 10 ⎦ ⎣0 0 2⎦
A = 3 6 −3 3 , rref(A) = 0 1 0 2 .
7 1

It's a matrix we row-reduced in Lecture 1.

1. interchange row R1 and row R2 :

⎡0 1 0⎤ ⎡0 1 0 ⎤⎡ 0 1 1 4 ⎤ ⎡ 3 6 −3 3 ⎤
⎣0 0 1⎦ ⎣0 0 1 ⎦⎣ −2 −3 10 ⎦ ⎣ −2 −3 10 ⎦
E12 = 1 0 0 , E12 A = 1 0 0 3 6 −3 3 = 0 1 1 4
7 7

2. multiply row R1 by 1 :
3

⎡3 0⎤ ⎡3 0 ⎤⎡ 3 3 ⎤ ⎡ 1 1 ⎤
1 1
0 0 6 −3 2 −1
E1 ( 13 ) = ⎢0 0 ⎥, E1 ( 13 )E12 A = ⎢0 0⎥ 0
⎣ −2 −3 10 ⎦ ⎣ −2 −3 10 ⎦
1 1 4 = 0 1 1 4
⎣0 1⎦ ⎣0 ⎦
1 1
0 0 1 7 7

3. add 2 times row R1 to row R3 :

⎡1 0 0⎤ ⎡1 0 0 ⎤⎡ 1 2 −1 1 ⎤ ⎡1 2 −1 1 ⎤
E13 (2)E1 ( 13 )E12 A =
⎣2 0 1⎦ ⎣2 0 1 ⎦⎣ −2 −3 10 ⎦ ⎣ 0 1 12 ⎦
E13 (2) = 0 1 0 , 0 1 0 0 1 1 4 = 0 1 1 4
7 5

4. subtract row R2 from row R3 :

⎡1 0 0⎤ ⎡1 0 0 ⎤⎡ 1 2 −1 1 ⎤ ⎡1 2 −1 1⎤
E23 (−1)E13 (2)E1 ( 13 )E12 A =
⎣ 0 −1 1⎦ ⎣ 0 −1 ⎦⎣ 12 ⎦ ⎣ 0 0 8⎦
E23 (−1) = 0 1 0 , 0 1 0 0 1 1 4 = 0 1 1 4
1 0 1 5 4

Notice that, except for the interchange Elelementary matrices Eij , the Elementary matrices given above are just
particular examples of the diagonal and triangular matrices having very simple inverses described earlier.
Can you see now how to continue the notation and definition to complete the row reduction and so illustrate the
following

Fundamental theorem: for every matrix A there exist Elementary matrices


E1 , E2 , … , Er−1 , Er so that

Er Er−1 … E2 E1 A = rref(A) .

You might also like