190 22. TOOLS USED IN PROOF OF PSEUDOLOCALITY
Suppose that there exists ti > l - ~aQ-^1 which is the first time, going
backwards in time from l, such that the closed ball Bg(f)(x, l 0 AQ-~) inter-
sects the boundary of Bg(ti) (xo, dg(f) (x, xo) + 190 AQ-~); we shall show that
this leads to a contradiction. Let x* be such a point of intersection, so that
9 - 1
(22.23) dg(t 1 )(x*,x 0 ) = dg(f)(x,xo) +
10
AQ-2.
Note that since x* E Bg(f)(x, 1 ~AQ-~), we have
1 - 1
(22.24) dg(f)(x*, x 0 ) :::; dg(f)(x, xo) +
10
AQ-2.
By the choice of ti and x* we have dg(t) (x*, xo) :::; dg(f) (x, xo) + { 0 AQ-^1!^2
for any t E [ti, l]. Hence if dg(t) (x, x*) :::; l 0 AQ-^112 , then
so that
(22.25)
(22.26)
dg(t) (x, xo) :S dg(t) (x, x*) + dg(t) (x*, xo)
9 - 1 1 - 1
:::; dg(f)(x,x 0 ) +
10
AQ-2 +
10
AQ-2,
Bg(t) (x*, 1 ~AQ-~) c Bg(t)(xo,dg(f)(x,xo) + AQ-~),
Bg(t) (xo,
1
1
0
AQ-~) c Bg(t)(xo,dg(f)(x,xo) + AQ-~)
for t E [t1, l] (with the second inclusion being obvious).
Note that by combining (22.9) and (22.21) to cover the cases where
(x, t) E Ma and (x, t) tj. Ma, respectively, we have
I Rm I (x, t) :S 4Q in Bg(t) (xo, dg(t) (x, xo) + AQ-~) x [t -~Q-^1 , t].
In particular, by (22.25) and (22.26),
Rc(x, t):::; 4nQ for x E Bg(t) ( x*, 1
1
0
AQ-~) U Bg(t) ( xo,
1
1
0
AQ-~)
and t E [ti, l]. Hence we may apply Theorem 18.7(2) on how distances
change under the Ricci fl.ow (with K = 4nQ and r 0 = lOfoQ-^112 )^5 to
conclude that
d 104 -1
dt dg(t) (x*, xo) ~ -5 (n - 1) vfnQ2 fort E [t 1 , ~.
(^5) This implies
-2(n-l)(Kra+r 01 )
= -2 (n - 1) ( 4nQ
10
~Q-^1 /^2 + 10ynQ1/2)
= _ 1~4 (n _ l) ynQ1;2_