The Inverse of a Partitioned Matrix

Herman J. Bierens
July 21, 2013

Consider a pair A, B of n n matrices, partitioned as

! !
A11 A12 B11 B12
A= , B= ,
A21 A22 B21 B22
where A11 and B11 are k k matrices. Suppose that A is nonsingular and
B = A1 . In this note it will be shown how to derive the Bij s in terms of
the Aij s, given that
det(A11 ) 6= 0 and det(A22 ) 6= 0. (1)
The latter conditions are sucient for the nonsingularity of A. However, in
general they are not necessary conditions. For example, consider the case
! !
A11 A12 0 1
A= = .
A21 A22 1 0
On the other hand, if A is positive definite then the conditions (1) are nec-
essary as well.
If B = A1 then
! !
A11 A12 B11 B12
AB = (2)
A21 A22 B21 B22
!
A11 B11 + A12 B21 A11 B12 + A12 B22
=
A21 B11 + A22 B21 A21 B12 + A22 B22
!
Ik Ok,nk
= ,
Onk,k Ink
where as usual I denotes the unit matrix and O a zero matrix, with sizes
indicated by the subscripts involved.

1
 To solve (2), we need to solve four matrix equations:

A11 B11 + A12 B21 = Ik (3)

A11 B12 + A12 B22 = Ok,nk (4)
A21 B11 + A22 B21 = Onk,k (5)
A21 B12 + A22 B22 = Ink (6)

It follows from (4) and (5) that

B12 = A1
11 A12 B22 , (7)
B21 = A1
22 A21 B11 , (8)

so that (3) and (6) become

A11 A12 A1
22 A21 B11 = Ik

A22 A21 A1
11 A12 B22 = Ink

Hence
1
B11 = A11 A12 A1
22 A21
1
B22 = A22 A21 A1
11 A12

Substituting these solutions in (7) and (8) it follows that

1
B12 = A1 1
11 A12 A22 A21 A11 A12
1
B21 = A1 1
22 A21 A11 A12 A22 A21

Thus,
1 1
A11 A12 A1
22 A21 A1 1
11 A12 A22 A21 A11 A12
A1 = 1 1
A1 1
22 A21 A11 A12 A22 A21 A22 A21 A1
11 A12

Moreover, since A.A1 = In implies A1 A = In , we also have

1 1
A11 A12 A1
22 A21 A11 A12 A1
22 A21 A12 A1
22
A1 = 1 1
A22 A21 A1
11 A12 A21 A1
11 A22 A21 A1
11 A12