352 26. BOUNDS FOR THE HEAT KERNEL FOR EVOLVING METRICS
Case 1. log (A j~~~D -fa~; :::::; -log 2. Then (26.59) implies
----BR^2 00
I (T) < e -zcor '""'2-i-l
R - f (T) L.J i=O
1 _ B_R2
= -_-e zc 0 r.
f (T)
Case 2. log(Aj~~~D-fc~; > -log2. Then by (26.56), (26.52), and
our assumption, we have
IR (T) :::::; in v^2 (x, T) dμ 9 ( 7 ) (x)
1
<---
- f (T)
2A BR2
< _ e -zc 0 r.
f (T/"f)
In either case we have
(26.60) { v^2 (x,T)dμ 9 ( 7 )(x)=JR(T):::; _
2
A exp(-B_R
2
)
ln-NR(K) f (T/'Y) 2CoT
usmg. f(T)^1 < f(Th)^1 < f(Th) 2A m. th e fi rs t case.
STEP 3. Exponentially weighted L^2 estimate. Given R > 0, let
Ko~ {x ED: d 9 (o) (x,K):::::; R},
Ki~ {x ED: 2i-lR:::::; d 9 (o) (x,K):::::; 2iR}
for i EN U {O}. We have for any D > 0 to be chosen sufficiently large later
r 2 d;(O) (x,K)
Jnv (x,T)e Dr dμ 9 ( 7 )(x)
1
2 d;(O) (x,!C)
= v (x, T) e Dr dμg(T) (x)
Ko
00 r 2 d;(O) (x,JC).
+ L }K v (x, T) e Dr dμg(T) (x).
i=l Ki
First note that
1
2 d~(O) (x,JC)
v (x, T) e Dr dμg(T) (x)
Ko
:::::; eDr R21 v^2 (x, T) dμg(T) (x)
Ko
1 R^2
< ---eDr
- f (T)