90 CHAPTER 4. MATRICES
Proof: From 4.16
BisAs j =(
0 Iri×ri 0)
B
0
Ips×ps
0
( 0 Ips×ps 0)
A
0
Iq j×q j
0
where here it is assumed Bis is ri× ps and As j is ps×q j. The product involves the sth blockin the ith row of blocks for B and the sth block in the jth column of A. Thus there are thesame number of rows above the Ips×ps as there are columns to the left of Ips×ps in those twoinside matrices. Then from Lemma 4.4.1 0
Ips×ps
0
( 0 Ips×ps 0)=
0 0 0
0 Ips×ps 0
0 0 0
Since the blocks of small identity matrices do not overlap,
∑s
0 0 0
0 Ips×ps 0
0 0 0
=
Ip1×p1 0
. . .
0 Ipp×pp
= I
and so ∑s BisAs j =
∑s
(0 Iri×ri 0
)B
0
Ips×ps
0
( 0 Ips×ps 0)
A
0
Iq j×q j
0
=(
0 Iri×ri 0)
B∑s
0
Ips×ps
0
( 0 Ips×ps 0)
A
0
Iq j×q j
0
=(
0 Iri×ri 0)
BIA
0
Iq j×q j
0
=(
0 Iri×ri 0)
BA
0
Iq j×q j
0
which equals the i jth block of BA. Hence the i jth block of BA equals the formal multipli-cation according to matrix multiplication, ∑s BisAs j. ■
Example 4.4.3 Let an n×n matrix have the form
A =
(a b
c P
)where P is n−1×n−1. Multiply it by
B =
(p q
r Q
)where B is also an n×n matrix and Q is n−1×n−1.
90 CHAPTER 4. MATRICESProof: From 4.160 0BisAsj = ( 0 Trix; 0 \B Ips xps ( 0 Ips xps 0 )A Iq; x4;0 0where here it is assumed Bis is rj X ps and As; is ps X gj. The product involves the s” blockin the i” row of blocks for B and the s“” block in the j“” column of A. Thus there are thesame number of rows above the Ip, x», as there are columns to the left of I,,p, in those twoinside matrices. Then from Lemma 4.4.10 0 0 0Tp, xps ( 0 Ip,xp, 9 )= O In,xp, 90 0 0 (0)Since the blocks of small identity matrices do not overlap,Ss0 0 0 0 Loy x Ppand so )), BisAs; =0 0 O In, xpy 0VY} 0 Iyxp, O |= =]0 0Y( O Tx, 0 )B Ip, xp, ( 0 Ip,xp, 0 \A Ig, x4;° 0 00 0=( 0 Tix; 0 jay Ips xps ( 0 Ips xps 0 )A Ig; x4;° 0 00 0=( 0 Six 0 )BIAY Iyyxgy [=( 0 Inxn 0 )BAL tayxa,0 0which equals the ij‘ block of BA. Hence the ij” block of BA equals the formal multipli-cation according to matrix multiplication, )°, BisAs;.Example 4.4.3 Let an n x n matrix have the form(oe)where Pisn—1xn—1. Multiply it by(75)where B is also ann Xn matrix and QO isn—1xn—1.