42 28. SPECIAL ANCIENT SOLUTIONS
estimate for Xo = 'Y (2kc-^1 l^2 ) , where 0::::: k::::: l zc:1;2 - ~J ~ f'..^1 That is,
e
Volgoo(t) Bgoo(t) ( 0 , r) ;:::: L Volgoo(t) Bgoo(t) ( 'Y ( 2kc-^1!^2 ) ) c-^1 /^2 )
k=O
;:::: (R + 1) /),c-n/^2
;:::: ( 2c=1/2 - ~) K,c-n/2. D
Partially motivated by Corollary 2 8. 13 , we have the following. Stronger moti-
vation comes from work of P erelman, which implies that the conjecture is true in
dimension n = 3; see Theorem 28. 20 b elow.
OPTIMISTIC CONJECTURE 28.14 (R and Vol of noncompact singularity mod-
els). Any noncompact singularity model defined on a time interval (-oo,w) must
have scalar curvature bounded from above on each interval (-oo, wo], where wo < w.
Consequently, th e volume growth of any noncompact singularity model must be at
least linear.
To b etter characterize when a singularity forms, it is n atural to consider the
following.
OPTIMISTIC CONJECTURE 28.15. For any finite-time singular solution g (t),
t E [O, T), on a closed manifold Mn , its scalar curvature must satisfy
sup R = oo.
Mx(O,T)
By the Hamilton-Ivey estimate (see Theorem 9.4 in Volume One) , this is
true when dim M = 3. Zhou Zhang [454] h as proved this conjecture for solutions
to the K a hler- Ricci fl.ow on closed manifolds.
We also have the following criterion for a singular solution to have unbounded
scalar curvature.
LEMMA 28.16 (Criterion for long-time existence under bounded scalar curva-
ture). Given a finite-time singular solution (Mn, g (t)), t E [O, T), on a closed man-
ifold, if there are no Ricci fiat singularity models associated to it, then sup M x [O,T) R
= 00.
PROOF. By Theorem 28.8, there exists an associated singularity model
(M~, 9 oo (t)), which is nonfl.at by definition. Suppose tha t supMx[O,T) R < oo.
Then , by the standard lower estimate for R , we have supMx[O,T) IRI < oo. This im-
plies that the scalar curvature of g 00 (t) is R 900 (t) = 0. By aRa7<'l = .6. 900 (t)Rg 00 (t)+
iRc 900 (t)J
2
, this implies Rc 900 (t) = 0, contradicting the hypothes is. D
A primary goal in studying singularity models is to understand their geometry
as well as their topology. For example, one anticipa t es the need to understand the
geometry at infinity of noncompact singularity models. The following m ay b e a
small step in this direction.
PROBLEM 28.17. What can one say about the asymptotic cones of noncompact
singularity models?
(^1) Her e, L · J : IR-+ Z denotes t h e floor function; i .e., LxJ is the greatest integer less than or
equa l to x. In particular, LxJ + 1 > x.