- THE COMPACTNESS THEOREM 231
for 0 :S: t :S: a/ K , where Cm~ J /3m (m - 1)! +a (Cm+ f3C:n) depends only
on m, n, and max {a, 1}. Thus we have
1 '7mR v m< I - a -tm <- --CmK tm/ 2
a
for 0 < t ::::; K.
This completes the inductive step, hence our proof of Theorem 7.1. D
- The Compactness Theorem
The Compactness Theorem for the Ricci fl.ow [64] tells us that any
sequence of complete solutions to the Ricci fl.ow having curvatures uniformly
bounded from above and below and injectivity radii uniformly bounded from
below contains a convergent subsequence. This result has its roots in the
convergence theory developed by Cheeger and Gromov. In many contexts
where this theory is applied, regularity is a crucial issue. By contrast, the
proof of the Compactness Theorem for the Ricci fl.ow is greatly aided by
the fact that sequences of solutions to the Ricci fl.ow enjoy excellent regu-
larity properties. (See Corollary 7.3 above.) Indeed, it is precisely because
bounds on the curvature of a solution to the Ricci fl.ow imply bounds on
all derivatives of the curvature that the compactness theorem produces C^00
convergence on compact sets. For this reason, we state it in this chapter.
We plan to present a detailed proof in a successor to this volume.
The most important application of the compactness theorem is the for-
mation of singularity models (final time limit flows). These are complete
nonfl.at solutions of the Ricci fl.ow that occur as limits of sequences of di-
lations about a singularity. Indeed, we already used it for this purpose in
Section 15 of Chapter 5. We shall discuss singularity models in greater
generality in Chapter 8.
THEOREM 7.8 (Compactness Theorem). Let
{Mi,gi (t), Oi, Fi: i EN}
be a sequence of complete solutions to the Ricci flow existing for t E (a, w),
where -oo ::::; a < 0 < w ::::; oo. Each solution is marked by an origin
Oi E Mi and a frame Fi = { ei, ... , e~} at Oi which is orthonormal with
respect to 9i ( 0). Suppose that there exists K < oo such that the sectional
curvatures of the sequence are uniformly bounded by K in the sense that
sup !sect [gi] I :S: K
Mjx(a,w)
for all i E N. Suppose further that there exists 6 > 0 such that the injectivity
radii of the sequence are bounded at Oi E Mi and t = 0 in the sense that
inj g;(O) ( Oi) 2: 6
for all i E N. Then there exists a subsequence which converges in the pointed
category to a complete solution
{M~, 900 (t), Ooo, Foo}