Universal covers and deck transformations

Assume X is path connected, locally path connected, and semilocally simply connected. The last condition means each point x has a neighborhood U such that the inclusion-induced map $\pi_1(U,x)\to\pi_1(X,x)$ is trivial, taking a suitable path-connected neighborhood when needed. Loops in U must contract in X, not necessarily inside U itself. Local simple connectivity is stronger. Under these hypotheses there exists a path-connected simply connected covering of X, unique up to the appropriate covering isomorphism after choosing a basepoint lift.

Semilocal simple connectivity is necessary for a simply connected cover: take an evenly covered neighborhood, lift a loop inside one sheet, contract it upstairs, and project that contraction. The shrinking wedge of circles, often called the Hawaiian earring, fails this condition at its accumulation point and does not have a classical simply connected covering of this kind. Path connectedness by itself cannot prevent that failure. Conversely, CW complexes have sufficient local regularity for the classical existence theorem, making them a reliable setting for the constructions in this module.

Fix $x_0\in X$ and take the points of the universal cover to be homotopy classes relative to endpoints of paths starting at $x_0$. Project a class to its terminal point. A small path-connected neighborhood whose loops die in X supplies neighborhoods upstairs by appending paths that stay in that neighborhood. Independence of these appended choices and the local sheet property use precisely the hypotheses above. It is not enough to make a set of path classes; the topology must be defined so that endpoint projection becomes a covering map.

A deck transformation is a homeomorphism $d:\widetilde X\to\widetilde X$ satisfying $pd=p$. In a connected cover, such a map is determined by the image of one point, using uniqueness of lifted paths, and a deck transformation fixing one point is the identity. For a universal cover the deck group acts freely and transitively on each fiber and is isomorphic to $\pi_1(X,x_0)$, with the conventional identification depending on path and multiplication choices. This action should not be confused with an arbitrary homeomorphism of the total space, which may move the projected point.

The universal cover of a circle is the real line and that of a torus is the plane, both contractible. These examples can mislead: the identity map of a two-sphere is its universal cover because the sphere is simply connected, yet the sphere is not contractible. Higher homotopy detects the surviving two-dimensional information. A universal cover removes π₁, not every homotopy group. Indeed, covering maps induce isomorphisms on higher homotopy groups, so those invariants persist upstairs rather than disappearing.

For a general connected covering corresponding to subgroup H of $G=\pi_1(X,x_0)$, the deck group is $N_G(H)/H$, where the normalizer consists of elements conjugating H to itself. Only when H is normal does this become the quotient G/H and act transitively on a fiber. A universal cover corresponds to the trivial subgroup, recovering the full group G. Thus the number of sheets does not in general equal the number of deck transformations. The next lesson will connect these statements with the pointed and unpointed classification of coverings.