Continuation and Breakdown Criteria

A local Einstein development can be continued past a finite time whenever the quantities required by the local existence theorem remain controlled and the foliation stays nondegenerate. The contrapositive is the useful breakdown statement. There is no universal theorem saying that one particular curvature scalar must diverge: continuation criteria depend on gauge, regularity, topology, matter, and the geometric control used to rebuild local coordinates.

Required background. Analysis for Mathematical Relativity supplies bootstrap and compactness estimates, The Einstein Equation supplies the evolution, PDEs and Green Functions for Gravitation supplies continuation for hyperbolic equations, and Tensor Fields and Natural Operations distinguishes geometric from coordinate control.

Helpful background. Causality and Global Structure explains extendibility and the domain of a foliation.

A local continuation theorem

Consider vacuum initial data $(\Sigma,h,K)$ in the standard metric regularity class

$$h\in H^s_{\mathrm{loc}}, \qquad K\in H^{s-1}_{\mathrm{loc}}, \qquad s>\frac52,$$

and evolve them in a fixed wave-gauge formulation on slices $\Sigma_t$, $0\le t\lt T$. For a uniform lifespan assume a compact slice, or a fixed atlas with uniformly controlled geometry and suitable global or weighted bounds on the perturbation of a reference metric. Local Sobolev regularity by itself gives no common time of existence at spatial infinity. A representative analytic continuation statement is:

If the metric remains uniformly Lorentzian, the slices remain uniformly spacelike, the gauge source remains regular, and the controlling $H^s\times H^{s-1}$ norm stays finite as $t\nearrow T$, then the reduced solution extends to $T+\varepsilon$. If the constraints and gauge constraint vanish, the extension is an Einstein development.

The proof is a restart argument. At each $t_n\nearrow T$, the local theorem supplies a lifespan $\varepsilon$ depending only on the common norm bound, chart control, and hyperbolicity constants, not on $n$. Choosing $T-t_n\lt\varepsilon$ produces a solution beyond $T$, and uniqueness patches it to the original development. Boundedness alone does not imply strong convergence of the data at $T$; the restart argument avoids needing that inference. The local existence and uniqueness framework is described in Friedrich and Rendall 2000, §§5.1–5.3.

Its contrapositive is precise but formulation-dependent: if the maximal wave-gauge solution ends at finite $T$, then a controlling Sobolev norm diverges or the gauge/slicing loses the nondegeneracy required by the theorem. It does not follow that the spacetime itself is singular. A bad coordinate map can make a regular metric’s components or lapse degenerate.

From coordinate norms to geometric control

Geometric criteria replace a large coordinate norm by invariant or foliation-covariant quantities from which suitable coordinates and Sobolev estimates can be recovered. Let $n$ be the future unit normal and

$$\pi_{\mu\nu}=\mathcal L_n g_{\mu\nu}$$

its deformation tensor. Using $K_{ij}=-h_i{}^\mu h_j{}^\nu\nabla_\mu n_\nu$, its orthonormal-frame components are $\pi_{ij}=-2K_{ij}$, $\pi_{ni}=D_i\log\alpha$, and $\pi_{nn}=0$, where $\alpha$ is the lapse. Thus $\pi$ controls both the deformation of the spatial geometry and the acceleration of its normal observers.

A precise example is the compact CMC theorem of Klainerman and Rodnianski 2010, Theorem 1.1 and equations (3)–(8). In their smooth vacuum setting, the compact slices have negative Yamabe type and are parametrized by $K=t\lt0$. The initial slice has a finite coordinate covering with uniformly comparable metric components and finite $L^2$ spacetime curvature. For an endpoint $t_*\lt0$, the bound

$$\sup_{t_0\le t\lt t_*} \left(\|K(t)\|_{L^\infty(\Sigma_t)} +\|D\log\alpha(t)\|_{L^\infty(\Sigma_t)}\right)\lt\infty$$

allows the spacetime and its CMC foliation to continue beyond $t_*$. Here the tensor norm $\|K(t)\|$ means the norm of $K_{ij}$, while the scalar CMC parameter is its trace. No smallness assumption is made. Curvature-flux and local-coordinate control are major conclusions established inside the proof, not extra bounds the reader may silently assume. The restriction $t_*\lt0$ matters: this theorem is not by itself a proof that the CMC foliation reaches $t=0$.

An asymptotically flat maximal foliation, with $K=0$, is a different setting. The paper discusses it but carries out the displayed main theorem in the compact CMC setting. Likewise, replacing the time supremum above by an $L^1_t$ bound requires a separate theorem. These versions should not be combined into one assertion.

The time norm matters as much as the field being bounded. A pointwise-in-time estimate and an integrated $L^1_tL^\infty_x$ estimate are not interchangeable, and curvature-flux control in $L^2$ does not imply a pointwise curvature bound. When quoting a criterion, the foliation interval and every spatial and temporal norm are part of the theorem.

Failure modes that must be separated

Gauge breakdown. The lapse may approach zero, the shift may become singular, or wave coordinates may form caustics while the underlying spacetime admits a regular chart.

Geometric degeneration. Curvature may concentrate, the injectivity or volume radius may collapse, or the induced metric may cease to be uniformly comparable to a reference geometry. A bounded scalar polynomial invariant alone does not exclude these failures.

Matter breakdown. For coupled fluids, a shock may destroy the classical matter solution before any vacuum-style curvature criterion fails. Vlasov or scalar matter has different continuation variables.

Boundary failure. With a timelike boundary, loss of compatibility or boundary regularity can terminate the initial-boundary value problem even when interior norms remain controlled.

End of the chosen foliation. A foliation can stop covering a regular extension. Continuation of that gauge is weaker than geometric extendibility, while continuation of the maximal globally hyperbolic development is stronger and requires a coordinate-independent construction.

Thus a valid singularity argument must identify which continuation theorem applies and prove that every way to satisfy its hypotheses fails. The later pages on inextendibility distinguish this from geodesic incompleteness and curvature blow-up.

Check your understanding

Apply the analytic criterion to Minkowski spacetime in standard inertial slices. Which quantities remain controlled, and what would a coordinate singularity prove?

Solution

Here $K_{ij}=0$, $\pi_{\mu\nu}=0$, the lapse is $1$, the shift is $0$, curvature vanishes, and every fixed-order Sobolev norm of the metric relative to the flat background is constant. The solution continues for all $t$. If another coordinate system becomes singular at finite coordinate time, that proves only failure of that chart or foliation; the inertial development supplies a regular extension.

The next step is not to evolve arbitrary data, but to construct data on the constraint surface. Begin with The Einstein Constraint Map.

References

• Friedrich, Helmut, and Alan D. Rendall. “The Cauchy Problem for the Einstein Equations.” In Einstein’s Field Equations and Their Physical Implications, Lecture Notes in Physics 540, 127–224. Springer, 2000. arXiv:gr-qc/0002074.
• Klainerman, Sergiu, and Igor Rodnianski. “On the Breakdown Criterion in General Relativity.” Journal of the American Mathematical Society 23 (2010): 345–382. arXiv:0801.1709.