Cosmic Topology

Einstein's equations tell us about the local curvature of spacetime, but they are completely silent on the global shape, the topology: of the universe. A flat universe ($k = 0$) could be the infinite plane $\mathbb{R}^3$, or it could be a three-torus $T^3$: a finite space where traveling far enough in any direction brings you back to your starting point. A positively curved universe ($k = +1$) could be the three-sphere $S^3$, or it could be a quotient space $S^3/\Gamma$ such as the Poincare dodecahedral space. Topology selection is a genuinely quantum-gravitational question, and it connects to some of the deepest unsolved problems in physics.

Topology vs. Curvature: A Fundamental Distinction

It is essential to understand that curvature and topology are independent properties. Curvature is a local quantity measured by the Riemann tensor; topology is a global property that describes how the space is connected. Two key examples illustrate the distinction:

• A flat sheet ($\mathbb{R}^2$) and a cylinder ($\mathbb{R} \times S^1$) have the same curvature (zero everywhere) but different topology. The cylinder has a nontrivial fundamental group $\pi_1 = \mathbb{Z}$
• A flat sheet and a flat torus ($T^2 = S^1 \times S^1$) again have zero curvature, but the torus is compact and finite, with $\pi_1 = \mathbb{Z} \times \mathbb{Z}$
• The three-sphere $S^3$ and the Poincare dodecahedral space $S^3/I^*$ have the same positive curvature, but the latter is 120 times smaller in volume

Einstein's field equations determine the metric (hence the curvature) but do not constrain the topology. Given any solution of Einstein's equations on a simply connected space, one can construct a solution on any quotient of that space by a freely acting discrete isometry group $\Gamma$, with exactly the same local geometry.

Characterizing Topology: Algebraic Invariants

The fundamental group $\pi_1$ is the most powerful topological invariant for our purposes. For $S^3/\Gamma$, we have $\pi_1 = \Gamma$. The fundamental group determines the deck transformations that identify points in the universal cover, and hence determines the eigenvalue spectrum of the Laplacian on the quotient space, which is directly observable through the CMB.

The Poincare Dodecahedral Space in Detail

The Poincare dodecahedral space $S^3/I^*$ is constructed by taking a regular spherical dodecahedron (a "curved" dodecahedron embedded in $S^3$) and identifying each pentagonal face with the opposite face by a $36°$ clockwise rotation (a $\pi/5$ twist). The result is a closed, orientable 3-manifold with:

• Fundamental group: $\pi_1 = I^*$, the binary icosahedral group of order 120
• Volume: $\text{Vol}(S^3/I^*) = \text{Vol}(S^3)/120 = 2\pi^2 R^3/120$ where $R$ is the curvature radius
• Homology: $H_1 = 0$ (the space is a homology sphere, indistinguishable from $S^3$ by homology alone)
• Curvature: Constant positive curvature $K = 1/R^2$, identical to $S^3$

Luminet, Weeks, Riazuelo, Lehoucq, and Uzan proposed this space as a cosmological model in 2003 after WMAP measured low power on large angular scales. The Laplacian spectrum on $S^3/I^*$ retains only selected modes of $S^3$, which can reduce low-multipole CMB power for a suitable topological size. Quantitative results still depend on the total density, observer position, and full radiation-transfer calculation. The model offers a possible explanation of the anomalies, not a confirmed parameter-free fit.

The Weeks Manifold

Among hyperbolic ($k = -1$) 3-manifolds, the Weeks manifold holds a special place: it is the closed hyperbolic 3-manifold of smallest known volume, $\text{Vol} \approx 0.9427$ (in units where the curvature radius is 1). Discovered by Jeffrey Weeks in 1985, it is constructed by identifying faces of a particular polyhedron in hyperbolic space. While positively curved spaces like $S^3/I^*$ require $\Omega_{\text{tot}} > 1$, negatively curved topologies require $\Omega_{\text{tot}} < 1$. Current observations constrain $\Omega_{\text{tot}}$ to be very close to 1, making hyperbolic topologies viable only if their fundamental domain is large compared to the curvature radius.

Why Topology Matters

The topology of space has observable consequences. In a finite universe, there is a maximum wavelength for any perturbation, modes larger than the space itself cannot exist. This produces specific signatures:

• Modified large-angle CMB: A compact topology with a sufficiently small scale discretizes modes and can reduce selected long-wavelength fluctuations
• Matched circles: If the universe is small enough, the CMB last-scattering surface wraps around, producing pairs of circles with matching temperature patterns
• Discrete spectrum: Quantum fields on a compact space have a discrete (rather than continuous) spectrum of allowed modes
• Casimir energy: The finite size generates a Casimir vacuum energy that depends on the specific topology

The Matched Circles Test

The matched circles test, proposed by Cornish, Spergel, and Starkman (1998), is the most direct observational probe of cosmic topology. The idea is elegant: if the universe is smaller than the diameter of the last scattering surface (LSS), then the LSS wraps around and intersects itself. The intersection of two copies of the LSS is a circle, and the temperature pattern along this circle must be the same when viewed from both copies (up to a possible phase shift from the face identification).

CMB Multipole Signatures

Beyond matched circles, nontrivial topology produces distinctive patterns in the CMB angular power spectrum $C_\ell$. In a multiply connected space, the eigenmodes of the Laplacian form a discrete set rather than a continuum, and the statistical isotropy of the CMB is broken in specific ways:

• Modified low multipoles: Discretizing the spectrum can reduce power in selected large-scale modes, but projection onto observed multipoles is not a simple term-by-term removal of $\ell=2$ and $\ell=3$
• Correlations between multipoles: In a simply connected universe, different $\ell$ modes are statistically independent. In a multiply connected space, the discrete mode spectrum introduces correlations between $C_\ell$ values at different $\ell$
• Violation of statistical isotropy: The preferred directions of the face identifications break rotational invariance, producing off-diagonal correlations in the $a_{\ell m}$ coefficients
• Platycosmic models: These are flat ($k=0$) compact spaces where one or more dimensions are much smaller than the others (e.g., a "slab" topology $\mathbb{R}^2 \times S^1$). These suppress only the modes along the short dimension, producing anisotropic large-angle correlations. They are called "platycosmic" (flat-universe) and are particularly interesting because they are consistent with $\Omega_{\text{tot}} = 1$ exactly

Current Observational Status

The observational situation remains tantalizing but inconclusive. The Planck 2018 results confirm several large-angle CMB anomalies that are consistent with (but not proof of) nontrivial topology:

• The quadrupole power is low at approximately the $5\text{-}10\%$ level (i.e., 90-95% of simulated $\Lambda$CDM universes have a higher quadrupole)
• The quadrupole and octupole are anomalously aligned (the "axis of evil"), which can arise naturally in certain topologies
• The lack of large-angle correlations: the two-point correlation function $C(\theta)$ is nearly zero for $\theta > 60°$, which is surprising in an infinite universe but expected in a compact one
• Constraints combining the CMB with cosmological distance data are consistent with zero spatial curvature. They still mathematically allow compact flat quotients and spherical or hyperbolic spaces with a very large curvature radius

The COMPACT program develops calculations and statistical methods for comparing compact topologies with cosmological data. The challenge is to compute full anisotropic correlations and then account for masks, foregrounds, and the look-elsewhere effect. No published result yet supports a detection of cosmic topology.

The Hartle-Hawking Wave Function and Topology

The no-boundary proposal of Hartle and Hawking defines the wave function of the universe as a Euclidean path integral over compact geometries. Applied to the question of topology, it makes a sharp prediction:

This is exponentially suppressed relative to the simply connected topology $S^3$, by the factor $\exp(-\pi(1 - 1/|\Gamma|)/G\Lambda)$. The Hartle-Hawking wave function overwhelmingly prefers $S^3$: the simplest topology.

The Sum Over Topologies

Should the gravitational path integral sum over all topologies? Hawking proposed yes (1978). But this runs into severe difficulties:

• 3D result (Carlip): The sum over topologies diverges, the density of topologies grows faster than the action suppresses them
• 4D undecidability: The classification of smooth 4-manifolds is algorithmically undecidable, making the sum not even constructively defined
• Exotic structures: In 4D, $\mathbb{R}^4$ has uncountably many inequivalent differentiable structures, all potentially contributing
• Recent progress (2026): Belin et al. identified surgery moves that generate the manifolds required in the 3D path integral, the first concrete framework for a controlled sum over topologies

Gibbons-Hawking Entropy and Topology

The de Sitter entropy for a quotient topology scales as:

A universe with topology $S^3/\Gamma$ has fewer microstates by a factor of $1/|\Gamma|$. For the Poincare dodecahedral space ($|\Gamma| = 120$), the Hilbert space dimension is $\exp(S_{\text{dS}}/120)$: vastly smaller than for $S^3$. Whether this means the quotient topology is less probable or that the universe simply has fewer observable degrees of freedom remains open.

Topology Change and Its Obstructions

Can the topology of space change over time? Classical GR places severe constraints:

Anderson and DeWitt showed that topology change in the presence of quantum fields produces infinite particle production. Causal dynamical triangulations (CDT) fix spatial topology by construction, and the resulting geometry has the correct semiclassical limit, suggesting topology may not fluctuate at all.

Observational Prospects

Current data (Planck, WMAP) do not detect nontrivial topology but do not rule it out for topological scales near the horizon. Active searches include:

• COMPACT collaboration: Searching for topology signatures in CMB data, including Casimir backreaction on expansion
• CMB-S4: Next-generation CMB telescope with the sensitivity to probe matched circles at larger angular scales
• DESI: BAO measurements that constrain the expansion history, which topology could modify

Counter-Arguments

• Topology selection may be unanswerable. The 4D classification problem is undecidable. The question may be ill-posed until a non-perturbative quantum gravity theory exists.
• CDT's fixed topology may be correct. If topology is a superselection sector rather than a dynamical variable, the question "why this topology?" has no dynamical answer.
• The question may be anthropic. If the landscape includes vacua with all spatial topologies, the observed topology may be an environmental accident.
• Inflation dilutes topology signals. Sufficient inflation pushes the topological scale far beyond the observable horizon, making the question empirically inaccessible.