Fiber-homotopy equivalence

From HandWiki

In algebraic topology, a fiber-homotopy equivalence is a map over a space B that has homotopy inverse over B (that is if ht is a homotopy between the two maps, ht is a map over B for t.) It is a relative analog of a homotopy equivalence between spaces.

Given maps p: DB, q: EB, if ƒ: DE is a fiber-homotopy equivalence, then for any b in B the restriction

f:p1(b)q1(b)

is a homotopy equivalence. If p, q are fibrations, this is always the case for homotopy equivalences by the next proposition.

Proposition — [1] Let p:DB,q:EB be fibrations. Then a map f:DE over B is a homotopy equivalence if and only if it is a fiber-homotopy equivalence.

There is also a completely analogous notion of a cofiber-homotopy equivalence, a homotopy equivalence such that the homotopies connecting the compositions and the identities are homotopies under the base. Then, similar to the above, we have:

Proposition — [2] Let i:AX,j:AY be cofibrations. Then a map f:XY under A is a homotopy equivalence if and only if it is a cofiber-homotopy equivalence.

Proof of the proposition

The following proof is based on the proof of Proposition in Ch. 6, § 5 of (May 1999). We write B for a homotopy over B.

We first note that it is enough to show that ƒ admits a left homotopy inverse over B. Indeed, if gfBid with g over B, then g is in particular a homotopy equivalence. Thus, g also admits a left homotopy inverse h over B and then formally we have hf; that is, fgBid.

Now, since ƒ is a homotopy equivalence, it has a homotopy inverse g. Since fgid, we have: pg=qfgq. Since p is a fibration, the homotopy pgq lifts to a homotopy from g to, say, g' that satisfies pg=q. Thus, we can assume g is over B. Then it suffices to show gƒ, which is now over B, has a left homotopy inverse over B since that would imply that ƒ has such a left inverse.

Therefore, the proof reduces to the situation where ƒ: DD is over B via p and fidD. Let ht be a homotopy from ƒ to idD. Then, since ph0=p and since p is a fibration, the homotopy pht lifts to a homotopy kt:idDk1; explicitly, we have pht=pkt. Note also k1 is over B.

We show k1 is a left homotopy inverse of ƒ over B. Let J:k1fh1=idD be the homotopy given as the composition of homotopies k1ff=h0idD. Then we can find a homotopy K from the homotopy pJ to the constant homotopy pk1=ph1. Since p is a fibration, we can lift K to, say, L. We can finish by going around the edge corresponding to J:

k1f=J0=L0,0BL0,1BL1,1BL1,0=J1=id.

References

  1. May 1999, Ch 7., § 5.
  2. May 1999, Ch 6., § 5.