This is a crosspost of this MSE question.

Let $A\overset{\alpha}{\rightarrow}B$ be a (Hurewicz) fibration.

  • The homotopy lifting property w.r.t a fiber $\alpha ^{-1}(b)$ furnishes for each path $b\to b^\prime$ in the base a continuous map $\alpha ^{-1}(b)\to \alpha ^{-1}(b^\prime)$. Moreover, this assignment extends to a functor $\pi_1B\longrightarrow \mathsf{hTop}$.
  • On the other hand, as a fibration $\begin{smallmatrix}A\\\downarrow\\B\end{smallmatrix}$ admits a Hurewicz connection $s$. Given such a connection it is tempting to send a path $b\overset{\gamma}{\to} b^\prime $ in the base to the following set function (analogously to covering space theory) $$\alpha^{-1}(b)\longrightarrow \alpha ^{-1}(b^\prime),\quad a\mapsto \operatorname{eval}_1s(a,\gamma).$$ I suspect this set function might be continuous, but I see no reason for it to be a homotopy equivalence, since $s(a,\gamma)$ need not be related in a nice way to lifts of opposite path $b\overset{\bar\gamma}{\leftarrow} b^\prime $.


  1. Is fiber transport along a Hurewicz connection continuous?
  2. (Assuming continuity) Is fiber transport along a Hurewicz connection functorial?
  3. (Assuming continuity) Does it coincide with the first transfer functor?
  4. Suppose the fibers are all homeomorphic. Are there any interesting conditions that make the transport functor $\pi_1B\longrightarrow \mathsf{hTop}$ lift to $\mathsf{Top}$? That is, can we obtain such homeomorphisms via transport?

I thought about possible functoriality of the transport along the Hurewicz connection. We want $\operatorname{eval}_1(a,\delta \ast \gamma)=\operatorname{eval}_1s(\operatorname{eval}_1s(a,\gamma),\delta,)$. We may consider the concatenation $s(\operatorname{eval}_1s(a,\gamma),\delta)\ast s(a,\gamma)$ which seems to lift $\delta \ast \gamma$, but I'm not quite sure where to go from here.


Let $p: E\to B$ be a map. Define $\Lambda(p) = E \times_B B^I$; this is the space of pairs $(x,\gamma)$ consisting of a point $x\in E$ and a path $\gamma: [0,1] \to B$ such that $\gamma(0) = p(x)$. There is an evident restriction map $$ \rho: E^I \to \Lambda(p) $$ where $E^I$ is the free path space of $E$.

The map $p: E\to B$ is a Hurewicz fibration iff $\rho$ has a section (easy exercise). A choice of (continuous) section $s: \Lambda(p) \to E^I$ defines transport.

Assume $p$ is a Hurewicz fibration. Consider the commutative diagram $\require{AMScd}$ $$ \begin{CD} \Lambda(p) \times 0 @>>> E \\ @VVV @VV pV \\ \Lambda(p) \times I @>>> B \end{CD} $$ where the bottom arrow is given by $((x,\gamma),t) \mapsto \gamma(t)$ and the top one is given by $((x,\gamma),0)\mapsto x$. By the lifting property, we can fill in the diagram with a map $$ \Lambda(p) \times I \to E \, . $$ This continuous map defines the Hurewicz connection by taking its adjoint to get a map $\Lambda(p) \to E^I$.

On the other hand, for a given path $\gamma$ wiht $\gamma(0) = b$ and $\gamma(1) = b'$, we can restrict the displayed map $\Lambda(p) \times I \to E$ to $F_b = (F_b \times_{\{b\}} \{\gamma\}) \times \{1\}$ (with $F_b = p^{-1}(b)$) to obtain a continuous map $$ F_b \to E $$ whose image is contained in $F_{b'}$. This map is your transport operation. This establishes (1)-(3).

As to (4), the transport functor restricts on closed based loops to a holonomy map $$ \Omega B \to G(F) $$ where $F$ is the fiber at the basepoint and $G(F)$ is the (group-like) topological monoid of self homotopy equivalences of $F$. This map is actually an $A_\infty$-homomorphism. By replacing $\Omega B$ with its Moore loops instead, this will become a morphism of topological monoids.

Let us make the additional hypothesis that your Hurewicz fibration is a fiber bundle (this is actually not a restriction up to fiber equivalence). Then your transport map gives a holonomy homomorphism $$ \Omega B \to H(F) $$ where $H(F)$ is the topological group of self homeomorphisms of $F$.

Your question (4) is more-or-less the condition that the connection admit a flat reduction, meaning this: let $H^\delta(F)$ be $H(F)$ with the discrete topology. Then we have a map of classifying spaces $$ BH^\delta(F) \to BH(F) $$ and (4) is answered in the affirmative iff the classifying map for the bundle $B\to BH(F)$ factors up to homotopy through $BH^\delta(F)$.

  • $\begingroup$ Dear John, I know being a Hurewicz fibration is equivalent to admitting a Hurewicz connection, but I don't understand how this affirmatively answers (1)-(3). $\endgroup$
    – Arrow
    May 3 '19 at 12:34
  • $\begingroup$ Let me explain differently. $\endgroup$
    – John Klein
    May 3 '19 at 13:10
  • $\begingroup$ The papers "Groupoid operations and fibre homoiopy equivalences I" link.springer.com/article/10.1007%2FBF01246620?LI=true and sequels should help. $\endgroup$ May 5 '19 at 17:34
  • $\begingroup$ Dear @JohnKlein, I am struggling with the holonomy map. The composite $\Omega B \to \pi_1(B,\ast)\to G(F)$ is connection independent. Assuming the fibration is a fiber bundle, how does a connection then define a map $\Omega B\to H(F)$? This doesn't seem to follow formally, and I'm not sure how to prove the parallel transports of a connection on a fiber bundle are homeomorphisms. I also don't see how to deduce this latter holonomy map respects multiplication. $\endgroup$
    – Arrow
    May 19 '19 at 13:28
  • $\begingroup$ Dear @JohnKlein, I have asked a more detailed version of this question on MSE. $\endgroup$
    – Arrow
    May 19 '19 at 23:25

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.