- LOCALIZED NO LOCAL COLLAPSING THEOREM
such that
f(u) ~I
1 if u E (-oo, !e^1 -n],
0::; ¢/(u)::; 12exp (en+l + n) if u E (!e^1 -n, ~e^1 -n),
exp (e1-~-u) if u E [~el-n,el-n),
oo if u E [e^1 -n,oo).
For any A> 0 there exists C1 (n, A) < oo such that
(18.66) ¢~u) (¢'(u))
2
- ¢"(u) 2: (2A + nen)<P'(u) - C1(n,A)¢(u)
for u E (-oo, e^1 -n), that is,
(18.67)
ituJ (¢'(u))^2 - ¢"(u) - (2A + nen)<P'(u)
¢(u) 2: -C1(n,A).
To see this, define
¢(u) (¢'(u))^2 - ¢"(u) - (2A + nen)<P'(u)
C3 (n) ~ uE[~el-n,%el-n] min ¢(u) < oo.
On the other hand, for u E rne^1 -n, e^1 -n) we have
(18.68)
ituJ (¢'(u))^2 - ¢"(u) - (2A + nen)<P'(u)
¢(u)
1 2 2A + nen
- (el-n - u) (^4) (el-n - u) (^3) (el-n - u) 2'
73
For u E rnel-n, e^1 -n) the RHS of (18.68) is bounded from below by some
finite number -C 2 (n, A). Therefore (18.67) holds for u E (-oo, e^1 -n) with
-C1(n, A)~ min {-C2(n, A), C3 (n)}.
The next step is to apply the weak maximum principle to the function
h: M x [O, 1] --+ [-oo, oo] defined by
h(w, t) ~ ¢(dg(t)(xo, w) - (2t - l)A) · (L(w, 1-t) + 2n + 1),
where L(w, T) = 4T£ (w, T).
Claim. Define
Then
(1)
where x is as in (18.60).
(2)
for T ::;! and w E M.
H (t) ~ min h (w, t).
wEM
H (1) = h (x, 1) = 2n + 1,
L(w, T) 2: -2n