148 30. TYPE I SINGULARITIES AND ANCIENT SOLUTIONS
STEP 3.
(30.57a)
(30.57b)
Define
The ODI f or u (s) in (30.48). By taking z = f3(s) in (30.56) we have
1
u'(s)::; ylW=tJA+Bu(s),
w-t
u (o) = e (q, t).
1 ( B s )
2
U(s)=:= B 2vw=t+JA+Be(q,t)
A
B '
which is the solution to the corresponding ODE
1
(30.58a ) U'(s)= ~JA+BU(s),
yw-t
(30.58b) u (o) = e (q, t).
By the ODE comparison theorem , we have u(s)::; U(s) for all s E [O,dg(t) (q,x)].
In particular, by takings= dg(t) (q, x), we obtain
f (xt)<_!_(Bdg(t)(q, x)+ IA+Be( t))2 A
' - B 2Jw - t v q, B
Bd~(t) (q, x ) A
::; 2 ( w - t) + B +^2 e ( q' t) '
which is (30.46).
STEP 4. Proof of (30.47). Taking q = p and using R (p, t) ::; :,
2
!~, we obtain
(using the constant path as the test path for e)
(30.59) f(p, t)::; ~ lw ~R (p , i ) di
2 w - t t
::; n^2 M.
Applying this to (30.46) yields (30.47). 0
We also have bounds for the first derivatives of e in both space and time.
COROLLARY 30.16 (Gradient and time derivative bounds fore of T ype I solu-
tions). Under the hypoth esis of Lemma 30.15, there exist constan ts C , D , E , F < oo
depending on n , c E (0, w2°'), and M such that
(30.60) IV£1 (x, t) ::; _C_ + D dg(t) (p , x)
ylW=t w-t
and
(30.61)
on M x [a + c, w).
l
ael() E Fd~(t)(p,x)
- xt <--+---~
at ' -w -t (w - t)^2
PROOF. (1) Applying (30. 4 7) to (30.56), we obtain
2A + 2Bn2 M B^2 d~(t) (p , x)
-----+ 2
w - t 2 (w -t)
(30.62) 1ve1 (x, t)::;
This proves (30.60) with C = J2A + 2Bn^2 M and D = ~' where A and Bare
given by (30.54)- (30.55).