PERSPECTIVITY BY DECOMPOSITION 29
« 0 + Co = n a 'n + n C 'm
n = l m—1
= fun <4 + 0 (by nil)
n = l m —1
oo oo
= n n k + c 'm ) (by nil)
n = l r a = l
= d.
Hence b0 fg aQ + c0. We have also a0b0 ^ ab = 0, cQbQ cb = 0. Finally,
if x is the axis of perspectivity of P(a, b), then a[ + x = b\ + x
(i = 1, 2, • • •), whence by III^ aQ x = b0 x. This, together with
a(p s^ax = 0, btfc ^ bx = 0, imphes Similarly 60~ c 0, and
Theorem 3.10 applies, yielding whence (a0, b0, c0)pd
by the Corollary to Definition 4.1. Combining this result with (10), (11),
(12), we may infer by Lemma 4.1, (a, b, c)pd.
Lemma 4.5: If and ab = 0, then {a, b, c)pd.
Pr o o f: Let b’ be an inverse of be in b. Thus
(13) b = be + b', b'c = b'bc = 0.
Let P(a, b) and P(b} c) be perspective isomorphisms of L(0, a), L(0, b)
and L(0, b), L(0, c) respectively, and let av c1 be the images of be under
P(a, b), P(b, c), respectively, and a2, c2 the images of b' under P(a, b),
P(b, c) respectively. By the Corollary to Theorem 3.3, c1 = be. Also,
av a2 are inverses in a, and cx, e2, i.e., be, e2, are inverses in c, whence
(14) ax + a2 = a, axa2 = 0, be + c2 = c
and (ax, be, bc)pd by Corollary 2 to Definition 4.1, since a±~ be. Now
a2<—' b'r^j c2, and a2b' ab = 0, b'c2 ^ b'c = 0, so that a2b' = b'c2 = 0.
Hence by Lemma 4.4, (a2, b', c2)pd, and the Corollary to Lemma 4.1
applies by (13) and (14), yielding (a, b, c)pd.
Lemma 4.6: If b ~ c and be = 0, then (a, b, c)pd.
Pr o o f: By hypothesis, b ~ a, cb = 0, whence Lemma 4.5 applies,
and (c, b, a)pd. Thus (a, b, c)pd by Corollary 3 to Definition 4.1.
Th e o r e m 4.1: If a ~ b ~ c , then (a, b, c)pd.
Pr o o f: This follows from Lemma 4.6 in precisely the same manner
as Lemma 4.5 follows from Lemma 4.4.
Le m m a 4.7: If (a i = 0, • • •, n) is a finite sequence such that
ai-1~ ai (i »), a0an = 0, then a0~ an.
Pr o o f: The lemma is obvious for n = 0, 1. Let m ^ 1 and suppose th at
it is true for » = 1, • • •, m. Then consider a sequence (a0, • • *, am+1) of