Universal Cover

The universal cover of a connected topological space is a simply connected space with a map that is a covering map. If is simply connected, i.e., is pathwise-connected and has a trivial fundamental group, then it is its own universal cover. For instance, the sphere is its own universal cover. For a connected, locally pathwise-connected topological space , the universal cover exists iff is semilocally simply connected. When it exists under these hypotheses, it is unique up to an isomorphism of covering spaces (Hatcher 2002, pp. 63-67). Local path connectedness is a hypothesis of this criterion, rather than a necessary condition for every simply connected covering space.

Geometric structures can be lifted using the local inverse charts of a covering map. For example, a Riemannian metric on a smooth manifold pulls back to a Riemannian metric on its universal cover, making the projection a local isometry. If this lifted Riemannian metric is flat and complete, then the universal cover is isometric to Euclidean space of the same dimension (Wilkins 2005, p. 63). The complex structure of a Riemann surface likewise lifts to its universal cover (McMullen 2026, p. 12). By the uniformization theorem, this universal cover is biholomorphic to the open unit disk, the complex plane , or the Riemann sphere (Voight 2021, pp. 642-643).

The above left diagram shows the universal cover of the torus, i.e., the plane. A fundamental domain, shaded orange, gives the torus after its opposite boundary edges are identified. The real projective plane is the set of lines through the origin, and its universal cover is the sphere, shown in the right figure above. The only nontrivial deck transformation is the antipodal map.

The compact Riemann surfaces with genera are -holed tori, and their universal covers are the unit disk. The octagon illustration above shows a hyperbolic regular octagon in the unit disk. With the colored edges identified, it is a fundamental domain for the double torus. Each hole has two loops, and cutting along each loop yields two edges per loop, or eight edges in total. Each loop is also shown in a different color, and arrows are drawn to provide instructions for lining them up. The fundamental domain is in gray and gives the double torus illustrated above after its paired boundary edges are identified. The above animation shows some translations of the fundamental domain by deck transformations, which form a Fuchsian group. They tile the unit disk by analogy with the square tiling the plane for the square torus.

Although it is difficult to visualize a hyperbolic regular octagon in the unit disk as a cut-up double torus, the illustration above attempts to portray this. It is unfortunate that no compact manifold surface of constant negative Gaussian curvature can be smoothly and isometrically embedded in . As a result, this picture is not isometric to the hyperbolic regular octagon. However, the generators for the fundamental group are drawn in the same colors, and are examples of so-called cuts of a Riemann surface.

For the torus and double torus shown above, a geometric construction begins by cutting the surface into a simply connected polygonal fundamental domain. The double torus gives the hyperbolic octagon, while the square torus gives a square. Copies of the fundamental domain are joined across paired edges. These identifications must reproduce the local neighborhood of every point of the original surface, including its polygon vertices. Continuing the identifications gives the plane or unit disk tiled by copies of the fundamental domain. This polygon construction describes these surface examples rather than an arbitrary topological space.

A general construction for connected, locally pathwise-connected, semilocally simply connected uses homotopy classes of paths beginning at a fixed basepoint . The homotopies keep both endpoints fixed. With the covering topology, the projection sends each homotopy class to the final endpoint of its path (Hatcher 2002, pp. 64-65).

When is locally pathwise-connected, every connected covering space of is itself covered by the universal cover of . In this sense, the universal cover is the largest possible covering space. More precisely, let be a covering map with connected, and let be the universal covering projection. Given points and with , there is a unique covering map satisfying and . This is the based universal property (Hatcher 2002, p. 68).

The Kollár-Pardon conjecture predicts that a connected normal projective complex variety has universal cover biholomorphic to a semialgebraic open subset of a projective variety exactly when the cover is . Here is a bounded symmetric domain, is a nonnegative integer, and is simply connected normal projective. Either or can be a point, and can be singular.