Timelike, Null, and Geodesic Completeness

Geodesic completeness asks whether freely falling trajectories can be continued for every value of their natural parameter. It is a property of the spacetime metric and manifold together, not a statement that curvature is bounded or that every observer lives forever. Timelike, null, spacelike, and all-geodesic completeness are distinct conditions; none may be substituted for another without a theorem.

Required background. Geodesics and Jacobi Fields supplies affine parametrization and maximal solutions; Curvature supplies the tidal quantities later compared with completeness. Helpful background. Lorentzian Metrics and Causal Cones supplies causal type; Causality and Global Structure orients the global role of incomplete curves; Exact Solutions of Einstein’s Equations supplies extension and counterexamples.

Affine extendibility

Let $\gamma:(a,b)\to M$ be a maximally extended geodesic with affine parameter $\lambda$. It is complete when $a=-\infty$ and $b=+\infty$. For a future-directed causal geodesic, future completeness means $b=+\infty$ and past completeness means $a=-\infty$; a spacelike geodesic has two parameter directions but no intrinsic future direction. A spacetime is timelike geodesically complete if every inextendible timelike geodesic is complete; define null and spacelike completeness analogously. Geodesically complete without a qualifier usually means all affinely parametrized geodesics of every causal type are complete.

The affine qualification is essential. If $\lambda$ is affine, then $A\lambda+B$ is affine for constants $A\neq0$ and $B$. Finite versus infinite endpoint is invariant under this change. An arbitrary monotone reparametrization can map a finite interval to the whole real line, but the new parameter is not affine and does not repair incompleteness.

For timelike geodesics parametrized by proper time $\tau$, the normalization

$$g(\dot\gamma,\dot\gamma)=-1$$

makes $\tau$ affine up to shift and orientation. Null geodesics have zero proper time, so an affine parameter must be chosen from the geodesic equation. Its scale is not canonical, yet completeness is. Spacelike geodesics can be parametrized by arclength, but their completeness addresses the semi-Riemannian geometry rather than an observer’s history.

Steinbauer formulates singularity theorems in precisely this language: their conclusion is causal geodesic incompleteness, not the presence of a singular point Steinbauer 2023, §2.3, pp. 8–9.

Inextendible curves and maximal parameters

An inextendible geodesic is one that has no continuation as a geodesic in the same spacetime beyond its parameter interval. Inextendibility of the curve must not be confused with inextendibility of the spacetime. A geodesic can stop at a finite missing boundary even when the spacetime has an obvious smooth extension.

Take the Minkowski half-space $M=\{t\lt0\}$. The curve

$$\gamma(\tau)=(\tau,0,0,0), \qquad -\infty\lt\tau\lt0,$$

is a future-inextendible timelike geodesic in $M$ and reaches its parameter endpoint after finite proper time. The restricted spacetime is timelike incomplete. Yet its curvature vanishes, and adjoining the surface $t=0$ inside full Minkowski spacetime gives a smooth extension in which the same geodesic continues. Completeness detects a defect of the chosen spacetime; it does not decide whether that defect is removable.

Ordinary differential-equation existence supplies a useful local check. If a geodesic remains in a compact subset of a smooth coordinate chart and its tangent remains bounded, its phase-space trajectory has an accumulation point and the geodesic equation can be continued. Thus a finite-affine-parameter maximal geodesic must escape every suitable compact subset of the tangent bundle. It may run toward a missing manifold endpoint, unbounded curvature, or divergent velocity in all regular charts. The theorem does not say which.

Timelike, null, and spacelike versions

The three completeness conditions probe different families of solutions to the geodesic equation. In a Riemannian manifold all nonzero tangents are of one type, and Hopf–Rinow links geodesic completeness to metric-space completeness. Lorentzian distance is neither positive definite nor a metric-space distance, so that equivalence does not transfer.

Logical implications between causal types require additional hypotheses. Null completeness controls light rays and appears in the Penrose theorem; timelike completeness controls freely falling massive ideal observers and appears in cosmological incompleteness results. Spacelike completeness concerns geodesics tangent to spacelike directions and can fail independently of causal completeness. Even “causal geodesic completeness,” meaning timelike and null completeness, leaves spacelike geodesics unaddressed.

There are constructed smooth Lorentzian examples separating timelike and null completeness. For an explicit example and its causal-type comparison, see Wald 1984, §9.1, pp. 215–216 and Figure 9.1. The indefinite geodesic flow can drive different norm classes toward different noncompact regions at different parameter rates. Exact symmetry or energy hypotheses can restore implications in special families, but no general statement permits replacing “future null incomplete” by “future timelike incomplete.” A theorem’s causal qualifier and time direction are part of its conclusion.

The distinction also matters under conformal rescaling. Unparametrized null geodesics are conformally invariant, but their affine parameters are not; null completeness can change. Timelike geodesic paths themselves generally change. A regular unphysical endpoint at finite coordinate therefore says nothing about physical completeness until the conformal factor has been integrated along the curve.

Worked example: infinite cosmic time, finite affine time

Consider the expanding flat patch of de Sitter spacetime,

$$ds^2=-dt^2+a(t)^2d\boldsymbol x^2, \qquad a(t)=e^{H_0t}, \qquad H_0\gt0,$$

with $t\in\mathbb R$ and $\boldsymbol x\in\mathbb R^3$. Here $H_0$ is constant and has units of inverse time. For a null geodesic moving only in the $x$ direction, translation symmetry conserves $P=a^2dx/d\lambda$, where $\lambda$ is affine. The null condition gives

$$\frac{dt}{d\lambda}=\frac{\lvert P\rvert}{a}, \qquad \lambda(t_0)-\lambda(-\infty) =\frac1{\lvert P\rvert}\int_{-\infty}^{t_0}a(t)\,dt =\frac{e^{H_0t_0}}{H_0\lvert P\rvert}.$$

Thus the past affine interval is finite even though $t$ runs to $-\infty$. The comoving timelike geodesics, by contrast, have $\tau=t$ and infinite past proper time. This does not make the patch timelike complete: timelike geodesics with nonzero conserved spatial momentum satisfy

$$\frac{d\tau}{dt}=\frac{a}{\sqrt{a^2+P^2}},$$

whose integral also converges as $t\to-\infty$. Testing one family of comoving observers misses these tilted geodesics. The patch is past null and timelike incomplete, although full de Sitter spacetime is geodesically complete. Its constant curvature remains finite, and its past boundary is removable in the global extension. Compare the global de Sitter geometry.

Completeness is not regularity

Completeness, curvature, and extendibility answer different questions:

• completeness asks whether a selected curve family has infinite affine range;
• curvature blowup asks whether specified tensor components, invariants, or tidal integrals diverge; and
• inextendibility asks whether the metric can be embedded properly into a larger spacetime with a declared regularity.

No pair is equivalent in general. The flat half-space is incomplete with zero curvature and is smoothly extendible. Conical or quasiregular examples can be incomplete with bounded curvature yet obstruct a smooth manifold filling under imposed topology. Conversely, a scalar curvature can become large along curves extending to infinite affine parameter, so curvature growth alone does not force incompleteness.

Geodesic completeness is also weaker than completeness of every accelerated causal curve. A timelike observer can accelerate toward infinity and accumulate finite proper time even in geodesically complete Minkowski spacetime. Singularity definitions based on all curves therefore specify an admissible family—often curves with bounded acceleration or a generalized-affine parameter—rather than assuming geodesic completeness settles them.

Exercise

On the manifold $M=\mathbb R\times(0,\infty)$ with metric $ds^2=-dt^2+dx^2$, consider (a) $\gamma(\tau)=(\tau,1)$ and (b) $\sigma(\lambda)=(0,\lambda)$ for $\lambda\gt0$. Classify their completeness and state what the result implies about curvature.

Solution

Curve (a) is a timelike geodesic defined for every $\tau\in\mathbb R$, so that particular geodesic is complete. Curve (b) is a spacelike geodesic whose affine parameter reaches $0$ in finite range while the missing point $x=0$ is not in $M$; it is incomplete. There are also timelike and null geodesics directed toward $x=0$ that reach it in finite affine parameter, so the spacetime is neither timelike nor null complete.

The Riemann tensor vanishes everywhere on $M$. The incompleteness therefore implies no curvature blowup. Embedding the half-plane into full two-dimensional Minkowski space smoothly continues the curves.

Incomplete Curves without Singular Points next explains what a finite-parameter curve endpoint can mean when it does not belong to the physical manifold.

References

• Beem, John K., Paul E. Ehrlich, and Kevin L. Easley, Global Lorentzian Geometry, 2nd ed., Chapter 6, Marcel Dekker, 1996.
• Steinbauer, Roland, “The Singularity Theorems of General Relativity and Their Low Regularity Extensions”, Jahresbericht der Deutschen Mathematiker-Vereinigung 125, §2.3, pp. 8–9, 2023.
• Wald, Robert M., General Relativity, University of Chicago Press, Chicago, 1984.