Achronal Boundaries and Null Generators

An achronal boundary is a closed causal frontier: no two of its points can be timelike related, although null-related points can lie on the same generator. Its low regularity is not an obstacle to causal reasoning. The boundary is a locally Lipschitz topological hypersurface, and the null generators of boundaries such as $\partial I^+(A)$ arise from the impossibility of replacing a boundary segment by a timelike shortcut. The precise generator statement requires more than the word “achronal,” so we keep its hypotheses visible.

Required background. We use timelike congruences and spatial projection to distinguish a family of curves from a hypersurface, and hypersurface normals and tangential projectors to interpret the smooth portions of a null boundary.

Helpful background. Pseudo-Riemannian metrics and signature fix the causal cones, affine and variational geodesics explain maximizing null segments, Causality and Global Structure provides orientation through this volume, and Exact Solutions of Einstein’s Equations supplies examples on which to test the definitions.

Future sets and their boundaries

Let $(M,g)$ be a connected, time-oriented spacetime with a $C^2$ Lorentzian metric. For the global statements below, $M$ has no spacetime boundary. A future set is a subset $F\subset M$ satisfying

$$I^+(F)\subset F.$$

Some authors reserve the name for open sets with $I^+(F)=F$. Nothing essential changes: the interior of any future set in the first sense is $I^+(F)$, and its boundary agrees with $\partial I^+(F)$. An achronal boundary is a set $H=\partial F$ for some future set $F$. Equivalently it can be presented as the boundary of a past set, with the time orientation reversed.

The basic example is $H=\partial I^+(A)$ for a subset $A$. The notation matters. $I^+(A)$ contains events reachable from $A$ by future-directed timelike curves, while $J^+(A)$ also admits null curves. These sets need not be equal or closed, but their relevant frontiers coincide under the standard construction. Replacing $I^+$ by a closure without explanation can conceal causal pathologies.

The boundary is achronal for a short set-theoretic reason. If $p,q\in H$ and $p\ll q$, then the openness of chronology and the future-set property force $q$ into $I^+(F)=\operatorname{Int}F$, contradicting $q\in\partial F$. Thus

$$I^+(H)\cap H=\varnothing.$$

This establishes achronality, not acausality: distinct points on one null generator are causally related. The future-set identities and the distinction between chronological, causal, and closed relations are developed in Minguzzi and Sánchez 2008, §2.4, pp. 10–12.

The achronal-boundary theorem

Under the assumptions above, every nonempty achronal boundary $H=\partial F$ has three robust properties:

1. $H$ is closed and is a locally Lipschitz embedded topological hypersurface.
2. Every timelike curve meets $H$ at most once, and $H$ has empty edge.
3. Every causal curve segment contained in $H$ is, up to reparametrization, a maximizing null geodesic segment. Two such segments that meet in their interiors either belong to the same geodesic or would create a forbidden timelike shortcut.

There is a stronger generator conclusion for causal wavefronts. Let $A$ be closed and take $H=\partial I^+(A)$. Through each $p\in H\setminus A$ passes a past-directed null generator contained in $H$. Followed toward the past, it either reaches $A$ or is past-inextendible in $M$. The time-dual statement holds for $\partial I^-(A)$. More generally, a future horizon is an achronal boundary whose generators have no future endpoints on the horizon.

The qualification “causal wavefront” is essential. An arbitrary achronal set need not be a hypersurface; even an arbitrary achronal boundary can contain points that are not causally related to any other boundary point. The null ruling through every point follows for the horizon or wavefront classes with the stated endpoint property, not from achronality alone. The complete topological and generator results, including the classification of possible endpoint points, are stated in Minguzzi 2019, §2.14, Theorem 2.87 and Lemma 2.89, pp. 59–63.

Why the theorem is true

The proof separates topology from null geometry.

First choose a small coordinate neighborhood $U$ of $p\in H$ and straighten a smooth timelike vector field there. We need controlled local cones, not causal convexity relative to the whole spacetime: arbitrarily small neighborhoods with that stronger property would require strong causality, which is not assumed. Integral curves sufficiently close to the one through $p$ run from the past side of the frontier into $F$. Each crosses $H$ exactly once by continuity and achronality. Consequently $H\cap U$, after shrinking the coordinate cylinder, is the graph of a function over a spacelike coordinate ball. The local light cones bound the slope of this graph, giving local Lipschitz continuity. Differentiability is neither assumed nor obtained everywhere.

The same two-sided crossing property proves that the edge is empty. Near any $p\in H$, every timelike curve from a nearby event on the past side to one on the future side must cross $H$. This is exactly what fails at the edge of a generic achronal set. A direct statement and proof appear in Minguzzi 2019, §2.16, Proposition 2.136, pp. 76–78.

For the generator step, approximate $p\in\partial I^+(A)$ by points of $I^+(A)$ and take causal curves from $A$ to those points. A limit-curve argument produces either a causal curve ending at $p$ from $A$ or a past-inextendible causal limit through $p$. No segment of the limit can be timelike: if it were, openness of $I^+$ would put $p$ inside $I^+(A)$. A causal segment on an achronal set is locally maximizing and hence is a null geodesic. If two generator pieces formed a genuine corner at an interior intersection, rounding the corner inside a convex neighborhood would yield a timelike curve between boundary points. That contradicts achronality and rules out branching in the interior.

This is proof architecture rather than a substitute for the limit-curve theorem. Its inputs are time orientation, local cone control and convex normal neighborhoods, closedness of the initial set for the stated endpoint alternative, and sufficient metric regularity for ordinary geodesic uniqueness. Convex normal neighborhoods control geodesics inside the neighborhood; they do not prohibit a causal curve from leaving and later re-entering it. At lower regularity, light-cone boundaries and generators require separate theorems.

Endpoints, caustics, and limitations

A generator can leave a future wavefront to the future after a focal point, a crossover, or another loss of achronality: beyond that point a timelike competitor exists. For a future event horizon the orientation is reversed in the important way—the generators have no future endpoints on the horizon, but they may have past endpoints where they enter it. Those entry points are where smooth-null-hypersurface intuition fails and are developed in Horizon Regularity, Caustics, and Crease Sets.

No curvature equation has been used. Flat Minkowski light cones already have a nondifferentiable vertex, so a corner or endpoint of an achronal boundary is not a curvature singularity. Conversely, a smooth null hypersurface need not be globally achronal; a null sheet can wrap around a spacetime or develop intersections. Nor does the theorem say that every horizon is an event horizon. The latter requires a specified asymptotic future and the global construction explained in Event Horizons as Global Boundaries.

Exercises

In four-dimensional Minkowski spacetime, let $A=\{(0,\mathbf 0)\}$. Determine $\partial I^+(A)$, identify its generators, and explain why its vertex does not contradict the local Lipschitz conclusion.

Solution

With $r=\lvert\mathbf x\rvert$, chronology gives $I^+(A)=\{t>r\}$. Hence

$$\partial I^+(A)=\{t=r,\ t\geq0\}.$$

For every unit spatial vector $\mathbf n$, the curve $\gamma_{\mathbf n}(\lambda)=(\lambda,\lambda\mathbf n)$ with $\lambda\geq0$ is an affinely parametrized future-directed null geodesic on the boundary. These rays are its generators. The graph $t=\lvert\mathbf x\rvert$ is Lipschitz because $\bigl\lvert\lvert\mathbf x\rvert-\lvert\mathbf y\rvert\bigr\rvert\leq\lvert\mathbf x-\mathbf y\rvert$, but it is not differentiable at $\mathbf x=0$. Thus the vertex demonstrates exactly why the theorem promises a Lipschitz topological hypersurface rather than a smooth null hypersurface.

Give an achronal boundary in Minkowski spacetime that has no null generators.

Solution

Take the future set $F=\{t\gt0\}$. Its boundary is the spacelike hypersurface $\{t=0\}$. Distinct points on it are spacelike separated, so no nonconstant causal curve lies in the boundary. This example satisfies the topological and Lipschitz conclusions but has no null ruling. In the wavefront theorem, choosing $A=\{t=0\}$ gives $H=A$, so the generator assertion for $H\setminus A$ is empty; there is no contradiction.

References

• Minguzzi, Ettore, and Miguel Sánchez. “The Causal Hierarchy of Spacetimes.” In Recent Developments in Pseudo-Riemannian Geometry, ESI Lectures in Mathematics and Physics, 299–358 (2008). arXiv.
• Minguzzi, Ettore. “Lorentzian Causality Theory.” Living Reviews in Relativity 22, article 3 (2019). DOI.