Continuation criteria and regularity questions

A local well-posedness theorem often gives a maximal interval and an alternative: either the solution continues, or its controlling regularity cannot stay bounded near a finite endpoint. The precise alternative depends on the local construction. For the one-dimensional subcritical H¹ NLS theory used earlier, a uniform H¹ bound permits a uniform restart and rules out a finite endpoint. At critical regularity the restart may depend on the profile's spacetime behavior, so a conserved norm alone may not supply that uniform interval.

For unforced three-dimensional Euler, take divergence-free whole-space initial data in H to the s with s strictly greater than five halves, and the corresponding classical solution on [0,T). The Beale–Kato–Majda theorem says a finite integral of the vorticity supremum up to T permits continuation beyond T in that regularity. Hence a finite maximal classical lifespan requires divergence of this integral. The theorem controls derivative amplification; kinetic energy alone does not bound that supremum integral. Periodic analogues exist, but the statement here uses the whole-space classical setting explicitly.

For unforced Navier–Stokes on $\mathbb R^3$, assume divergence-free initial data in $H^s$, $s>5/2$, and the solution $u\in C([0,T);H^s)$, smooth at positive times. A standard nonendpoint Prodi–Serrin continuation criterion requires velocity in L to the p time, L to the q space with q strictly greater than three, p at least two, and two over p plus three over q at most one. A finite such norm up to a finite maximal time prevents breakdown there. This is a conditional statement about the specified solution, not a bound that the energy inequality already provides.

For example, p=4 and q=6 gives equality and a scale-invariant criterion. The energy class only directly supplies p=2,q=6 through Sobolev embedding, whose sum is three halves rather than one. Temporal integrability is missing. The q=3,p=infinity endpoint is excluded from this stated elementary version; a deeper endpoint theorem is known in the standard whole-space theory and requires a separate proof. Excluding it here does not claim it is false. Any forced or bounded-domain version must include its own force and boundary hypotheses.

A finite value in one of these norms proves continuation under the theorem's hypotheses. A divergent upper bound proves nothing about actual divergence, and even actual divergence of a sufficient control norm need not prove a singularity without the converse theorem. Energy estimates, simulations with increasing gradients, and dimensional scaling each give different kinds of evidence. A proof of global regularity would need the required uniform control for its data class, or another argument; a breakdown construction must verify that the full equation and its stated regularity assumptions hold up to the proposed singular time.

Research-status note, checked October 18, 2026: Clay's September 11 notice reported an apparent-settlement announcement concerning its Navier–Stokes problem and described a deliberately unhurried evaluation process. This is not a prize-award statement or a proof supplied by this lesson. Fefferman's original formulation distinguishes whole-space and periodic alternatives, with unforced global-smoothness assertions and breakdown alternatives allowing specified smooth forcing. An announced result must be matched to the actual alternative and hypotheses before drawing conclusions about unforced global regularity. The classical continuation and weak-existence theorems above remain separate mathematical statements throughout that evaluation.