🔢 mathematics

Topological Obstructions to Riemannian-Lorentzian Continuation

This paper establishes that a smooth manifold can be partitioned into Riemannian and Lorentzian regions connected by a signature-changing hypersurface if and only if a specific relative Euler class obstruction vanishes, thereby defining precise topological selection rules for such geometro-topological continuations.

Original authors: Nathalie E. Rieger, Ödül Tetik

Published 2026-09-24

📖 7 min read🧠 Deep dive

CC BY 4.0

Original authors: Nathalie E. Rieger, Ödül Tetik

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine the universe not as a static stage, but as a shape that can change its very nature. In the study of the cosmos, scientists often try to understand how the universe began. One popular idea suggests that before time and space as we know them existed, the universe was in a different state, one where time did not flow forward and backward but behaved more like a fourth direction of space. This is called a Riemannian state. Eventually, this state would have to shift into the Lorentzian state we live in today, where time moves in a single direction and light cones define what can influence what. The big question is: can these two very different shapes of reality be stitched together smoothly? If they cannot, then the idea of a universe starting from a timeless, space-like beginning might be mathematically impossible, regardless of how elegant the theory sounds.

A team of mathematicians has now mapped out the strict rules that govern whether such a stitch is possible. They studied the geometry of these transitions, looking for the hidden topological barriers that might prevent a smooth change from a timeless, space-like region into a time-filled universe. Their work does not solve the physics of the Big Bang, nor does it tell us which specific model of the early universe is correct. Instead, it provides a set of geometric selection rules. Think of it as a blueprint check: before you can build a house, you must ensure the foundation can support the weight. Similarly, before a universe can transition from a timeless beginning to a time-filled existence, its shape must satisfy specific mathematical conditions. If the shape fails these conditions, the transition simply cannot happen, no matter how you try to arrange the physics.

The researchers focused on a specific way of joining these two regions. They imagined a smooth surface, like a membrane, separating a region where the geometry is purely space-like from a region where it includes time. For the transition to be physically meaningful, this membrane must not be a jagged tear or a singularity. Instead, the change must happen transversely, meaning the geometry shifts smoothly across the boundary. The team discovered that for this smooth transition to exist, the region representing our time-filled universe must possess a specific kind of vector field—a mathematical arrow that points everywhere in a consistent direction without ever stopping or vanishing. In the language of geometry, this requirement is tied to a quantity called the Euler characteristic, which is a number that describes the overall shape of a space, counting its holes and bumps in a specific way.

The main finding is that the possibility of this transition depends entirely on the shape of the time-filled region and how the boundary arrows point. If the time-filled region is compact, meaning it is finite and closed off like a sphere, the transition is only possible if the number of "bangs" (where time begins) and "crunches" (where time ends) on the boundary are balanced in a very precise way relative to the shape's Euler characteristic. If the time-filled region is infinite, the rules are different. For instance, if the universe is infinite in both the time-less and time-filled directions, the transition is only possible if the entire manifold has at least two distinct "ends" or directions in which it stretches out to infinity. If the universe has only one end, like a flat plane extending forever in all directions, it cannot undergo this specific type of smooth transition with a compact boundary.

The paper also clarifies what happens when the universe changes its topology, or its connectivity, during this process. In some scenarios, a single connected region might split into two separate regions as time begins. The researchers showed that if the time-filled region is to have a well-behaved causal structure—meaning time flows logically without loops or paradoxes—then the transition must happen in a way that avoids creating these topological changes within the time-filled region itself. If the topology changes, it must happen entirely within the time-less, space-like region before the transition occurs. This places a strong constraint on models of the early universe that rely on the universe splitting or merging as it comes into existence.

One of the most striking results concerns the dimension of the universe. The rules change depending on whether the total number of dimensions is even or odd. In an even-dimensional universe, such as our familiar four dimensions of space and time, the time-filled region must have a specific topological property: its Euler characteristic must be zero. If it is not, the transition is impossible. In odd dimensions, the rules are slightly more flexible, allowing for a balance between the starting and ending points of time on the boundary. This means that for a four-dimensional universe, the shape of the time-filled region is severely restricted. It cannot be just any shape; it must be a shape that allows for a consistent flow of time from start to finish without topological obstructions.

The researchers also explored cases where the transition surface itself is infinite. They found that if the boundary between the time-less and time-filled regions is infinite, the rules are much more permissive. In these cases, the transition can occur on almost any open manifold, provided the geometry can be constructed to allow the necessary vector field to exist. This suggests that while a finite, closed universe has very strict requirements for a smooth beginning, an infinite universe has more freedom to undergo such a transition. However, even in these infinite cases, the geometry must still be constructed carefully to ensure the transition is smooth and does not introduce singularities.

Ultimately, this work serves as a filter for cosmological theories. It does not tell us which theory of the universe's origin is true, but it tells us which theories are geometrically impossible. If a proposed model of the early universe requires a transition that violates these topological rules, that model cannot be realized as a smooth signature-changing manifold. The results act as a necessary condition: any viable candidate for a gravitational instanton, a theoretical object that describes the birth of a universe, must pass these geometric tests. If the relative Euler class obstruction is not zero, or if the compactness types do not match the allowed configurations, the transition simply cannot exist.

The study also highlights the difference between a universe that is finite and one that is infinite. For a finite universe to begin, the boundary where time starts must be carefully arranged. If the universe is finite, the transition surface must be compact, and the time-filled region must satisfy the Euler characteristic condition. If the universe is infinite, the transition surface can be compact or infinite, but if it is compact, the universe must have at least two ends. These distinctions are crucial for understanding the global structure of spacetime. The researchers constructed explicit examples for each allowed case, showing that these configurations are not just theoretical possibilities but can be mathematically realized.

In summary, the paper establishes that the smooth transition from a timeless, space-like beginning to a time-filled universe is not a free choice. It is constrained by the deep topology of the space itself. The shape of the universe, the number of its dimensions, and the way its boundaries are arranged all dictate whether such a transition can occur. For a universe to emerge smoothly from a timeless state, its geometry must be just right. If the shape is wrong, the transition is blocked by a topological obstruction, and the universe cannot begin in the way the theory proposes. This provides a rigorous mathematical foundation for evaluating models of the early universe, ensuring that only those consistent with the fundamental geometry of spacetime are considered viable.

Technical Summary: Topological Obstructions to Riemannian–Lorentzian Continuation

Problem Statement
The paper addresses the global and topological question of when a Riemannian manifold region ( $M_R$ ) can be smoothly continued into a Lorentzian region ( $M_L$ ) through a hypersurface ( $H$ ) where the metric signature changes transversely. Motivated by instanton continuation problems in quantum cosmology (specifically the Hartle–Hawking no-boundary proposal) and models of signature change in loop quantum cosmology, the authors investigate the existence of a smooth, transverse type-changing metric.

The central question is formulated as follows: Given a smooth manifold $M$ decomposed into $M = M_R \cup_H M_L$ , under what conditions does there exist a metric $\tilde{g}$ that is Riemannian on $M_R$ , Lorentzian of index one on $M_L$ , and degenerate of corank one exactly along $H$ , with a transverse radical?

Methodology and Ansatz
The authors adopt a specific ansatz for the signature-changing metric, inspired by constructions in semi-Riemannian geometry:
$\tilde{g} = g - f V^\flat \otimes V^\flat$
where:

$g$ is an auxiliary Riemannian metric on $M$ .
$f \in C^\infty(M)$ is a smooth scalar function determining the signature change locus ( $H = \{f=1\}$ ).
$V \in \Gamma(TM)$ is a smooth vector field determining the direction of the signature change.
$V^\flat = g(V, \cdot)$ .

The problem is reduced to finding a triple $(g, f, V)$ satisfying specific geometric conditions:

$f$ must define a "collar-gluing" where $M_R = \{f < 1\}$ , $H = \{f = 1\}$ , and $M_L = \{f > 1\}$ , with $df \neq 0$ on $H$ .
$V$ must be nowhere vanishing on $M_L \cup H$ and transverse to $H$ .
The metric $\tilde{g}$ must be Riemannian on $M_R$ and Lorentzian on $M_L$ .

The authors demonstrate that the existence of such a metric is equivalent to the existence of a nowhere-vanishing vector field on $M_L$ that is transverse to the boundary $H$ . This transforms the geometric problem into a topological obstruction theory problem involving the relative Euler class.

Key Contributions and Results

1. The Main Theorem (Admissibility Criterion)
The primary result establishes that a collar-gluing $M = M_R \cup_H M_L$ admits an admissible Riemannian–Lorentzian continuation if and only if the Lorentzian side $M_L$ admits a smooth, nowhere-vanishing vector field whose restriction to $H$ is transverse to $H$ .

In terms of obstruction theory, this condition is equivalent to the vanishing of the relative Euler class:
$e(T M_L, \nu) \in H^n(M_L, H; \mathbb{Z}_{w_1}) = 0$
where $\nu$ is the prescribed transverse boundary field on $H$ .

Noncompact $M_L$ : If every connected component of $M_L$ is noncompact, the obstruction vanishes automatically (due to Lefschetz duality).
Compact $M_L$ : If $M_L$ has compact components, the obstruction is equivalent to the condition:
$\chi(X) = \chi(\partial_- X)$
for every compact connected component $X \subset M_L$ , where $\partial_- X$ denotes the union of boundary components where the vector field points into $X$ . In even dimensions, this simplifies to $\chi(X) = 0$ .

2. Geometro-Topological Selection Rules
The paper classifies all possible admissible configurations based on the compactness of $M_R$ , $M_L$ , and $H$ . The authors prove that exactly five types of configurations exist, with specific admissibility conditions:

(A) Compact $M_R$ , Noncompact $M_L$ , Compact $H$ : Always admissible on any connected open manifold. This corresponds to the standard "no-boundary" scenario.
(B) Compact $M_R$ , Compact $M_L$ , Compact $H$ : Admissible if and only if the relative Euler class of $M_L$ vanishes (i.e., $\chi(M_L) = \chi(\partial_- M_L)$ ).
(C) Noncompact $M_R$ , Noncompact $M_L$ , Compact $H$ : Admissible if and only if $M$ has at least two ends.
(D) Noncompact $M_R$ , Noncompact $M_L$ , Noncompact $H$ : Always admissible. This case requires "ad hoc" constructions (e.g., using properly embedded rays) rather than global Morse functions, as proper functions cannot yield noncompact level sets.
(E) Noncompact $M_R$ , Compact $M_L$ , Compact $H$ : Admissible if and only if the relative Euler class of $M_L$ vanishes.

3. Topological Constraints on the Riemannian Region
The authors derive constraints on the Riemannian region $M_R$ that arise from the topology of the whole manifold $M$ and the Lorentzian side $M_L$ . Using the Mayer-Vietoris sequence, they show:
$\chi(M_R) + \chi(M_L) = \chi(M) + \chi(H)$

If $\dim M$ is even and the gluing is admissible, then $\chi(M_R) = \chi(M)$ (since $\chi(M_L)=0$ and $\chi(H)=0$ ).
If $\dim M$ is odd, $\chi(M)=0$ , implying $\chi(M_R) = \chi(M_L)$ .

4. Examples and Physical Interpretation

Big Bang vs. Big Crunch: The orientation of the vector field $V$ relative to $H$ distinguishes between "big bang" (initial) and "big crunch" (final) transitions. The relative Euler class condition dictates how these transitions must be distributed on the boundary components of a compact Lorentzian region.
Sphere Example ( $S^n$ ): For $M=S^n$ with two spherical caps ( $M_R$ ) and a cylindrical band ( $M_L$ ), the authors show that for odd $n$ , a configuration with two big bangs is obstructed, while a "bang then crunch" configuration is admissible.
Topology Change: The paper notes that while topology change can occur, it is restricted. If $M$ is the interior of a compact manifold with boundary, the Lorentzian region can be diffeomorphic to $H \times [0, \infty)$ , implying no topology change occurs within the Lorentzian era itself if the transition hypersurface is chosen appropriately (Corollary 4.1).

Significance and Claims
The authors explicitly state that their work provides necessary geometric-topological selection rules, not sufficient conditions for a physical instanton.

Scope: The results establish the "geometro-topological foundation" required to pose questions about instanton continuation. They do not solve the field equations (Einstein or otherwise) or address analytic regularity beyond the smoothness of the metric tensor.
Limitations: The paper does not claim to replace or generalize specific Wick rotation constructions that rely on high symmetry or complex analytic continuation. Instead, it offers a rigorous topological filter: if the relative Euler class of the intended Lorentzian region does not vanish (or if the compactness/end conditions are not met), no such signature-changing metric of the specified type exists.
Implication: Consequently, a Riemannian manifold that fails these topological criteria cannot serve as a precursor to a Lorentzian universe within this specific framework of transverse signature change.

The paper concludes that the existence of a transverse Riemannian/Lorentzian type-changing metric is topologically equivalent to presenting the manifold as a collar-gluing of one of the five types (A–E) described, subject to the Euler class obstructions.

Technical Summary: Topological Obstructions to Riemannian–Lorentzian Continuation

More like this