8.01 Lifting criterion

Below the video you will find accompanying notes and some pre-class questions.

Notes

Lifting criterion

(0.00) In this section, we will see a very general result which tells us when we can lift a map to a covering space. This generalises path- and homotopy-lifting.

(0.23) Let p:YX be a covering space and let T be a path-connected, locally path-connected space. Suppose there is a continuous map f:TX. Given a point tT (with f(t)=x) and a point yp1(x), there exists a unique continuous lift ˜f:TY with ˜f(t)=y if and only if fπ1(T,t)pπ1(Y,y).
(3.44) If there is a lift ˜f then the criterion holds because for any loop γ in T based at t, f[γ]=(p˜f)[γ]=p(˜f[γ])pπ1(Y,y),
as ˜fγ is based at y.

(4.27) The converse is harder: if the criterion holds then we need to construct ˜f. We require ˜f(t)=y, which is a start, but we need to extend this to a map ˜f:TY. Given another point tT, pick a path α from t to t (we can do this because T is path-connected). Restrict f to the path α. This yields a path fα in X, so path-lifting gives us a path ~fα:[0,1]Y with initial condition y. We will define ˜f(t)=~fα(1).

(6.22) We need to check:

(6.57) The fact that ˜f is a lift of f is clear from the construction: the path ~fα is a lift of fα, so p(˜f(t))=p(~fα(1))=f(α(1))=f(t).

(7.40) The uniqueness of ˜f follows from the uniqueness lemma for lifts which we proved in an earlier section.

(8.01) To see that ˜f is well-defined, pick two different paths α1 and α2 from t to t. Let ˜f1(t)=~fα1(1) and ˜f2(t)=~fα2(1). We need to check that ˜f1(t)=˜f2(t).

Note that f(α1(1))=f(α2(2))=f(t). The loop γ:=α12α1 is a loop in T based at t and its image under f is a loop in X based at x. By assumption, f[γ]pπ1(Y,y), which means that there is a loop δ in Y based at y such that pδ=fγ. Now I claim that ˜f1(t)=δ(1/2)=˜f2(t).

(11.12) To see that ˜f1(t)=δ(1/2), note that δ|[0,1/2] is a lift of fα1 starting at y, so δ(1/2)=~fα1(1)=˜f1(t). Similarly, δ1|[0,1/2] is a lift of fα2 starting at y. Therefore δ1(1/2)=~fα2, but δ1(1/2)=δ(1/2) so ˜f2(t)=~fα2(1)=δ(1/2).

(12.22) Finally, we need to show that ˜f is continuous. Take an open set VY; we want to show that ˜f1(V) is open. In fact, it's sufficient to check this for a V running over a base of the topology of Y. We will take the base comprising the elementary sheets over elementary neighbourhoods of the covering map.

(14.50) Pick a point tT in ˜f1(V). The point f(t) then lives in an elementary neighbourhood UX. We need to find an open set WT containing t such that ˜f(W)V. Take f1(U)T; this is open because f is continuous and let Wf1(U) be a path-connected open subset of f1(U) containing t; I can do this because T is locally path-connected (every open neighbourhood of a point t contains a path-connected open neighbourhood of t).

(17.08) To show that such a W satisfies ˜f(W)V, pick tW and a path α in W from t to t. To define ˜f(t), we simply take ~fα(1), where ~fα is the unique lift of fα starting at ˜f(t). This lift is given by qfα, where q:UV is the local inverse for the covering map over the elementary neighbourhood U. This implies that ˜f(t)=q(f(α(1)))V, so ˜f1(V) contains W. This proves that ˜f is continuous.

(19.32) The video then recaps the proof, because it is quite complicated.

Applications

(21.10) Any continuous map F:S2T2 is nullhomotopic.
Consider the covering map p:R2T2 coming from the properly discontinuous action of Z2 on R2. Since π1(S2,t)={1} for any basepoint tS2, any continuous map F:S2T2 satisfies Fπ1(S2,t)={1}π1(T2,F(t)). For any yp1(F(t)), we have pπ1(R2,y)={1}, so Fπ1(S1,t)pπ1(R2,y),
and the map F lifts to a map ˜F:S2R2. This lift is nullhomotopic because R2 is contractible. Let H be such a nullhomotopy. Then pH is a nullhomotopy of F.

The set of based homotopy classes of maps from S2 into another space X is called π2(X); this theorem shows that π2(T2) is trivial. More generally, the higher homotopy groups of a space X are the groups πn(X) of based homotopy classes of maps SnX. The same reasoning shows that πm(X) is trivial for all m2 whenever X admits a contractible covering space.

Pre-class questions

  1. Can you think of another space with trivial π2(X)?
  2. I claim that any covering space of a CW complex is a CW complex. How would I construct the cells of this covering space?

Navigation