By H-space I mean a unital magma in $Ho(\mathsf{Top}_*)$, i.e. the homotopy category of pointed spaces (feel free to use any of your favourite model :) ).

This means that we have a pointed space $X$ and a map $\mu_X:X \times X \to X$ such that $\mu_X \circ j \simeq \nabla$, where $\nabla : X \vee X \to X$ is the fold (or co-diagonal) map, and $j:X \vee X \to X \times X$ is the pushout-product of the map $* \to X$ with itself.

An H-map between H-spaces is a map $f:X \to Y$ of the underlying spaces such that the obvious square involving $f$ and the multiplication maps is homotopy commutative.

Here comes the question: how to prove that the homotopy fibre of $f$ is again an H-space (in a sensible way, of course)?

Homotopy pullbacks of spaces are weak pullbacks in the homotopy category, and that's how you get a multiplication on the fibre compatible with that of $X$. However, I can't seem to find how to check the desired (homotopy) commutativity, as that would seem to require that weak pullback to be a strict one, which is known to be false.

Thanks in advance for any answer/comment!

  • $\begingroup$ Do you know this to be true? $\endgroup$
    – Mark Grant
    Jan 26, 2017 at 11:23
  • $\begingroup$ It's claimed in lots of places (without proof), though I am starting to question its validity $\endgroup$ Jan 26, 2017 at 11:26
  • $\begingroup$ Is there something wrong with this argument? Replace the map f with a homotopy equivalent fibration. This should still be a map of H-spaces, and now the actual fiber is the homotopy fiber and also an H-space. It's homotopy equivalent to the homotopy fiber of the original map, whence the claim. $\endgroup$ Jan 26, 2017 at 16:31
  • $\begingroup$ The fact that the fibre of the replaced map is an H-space is exactly what I am trying to prove, but I can't get unitality under my assumptions (I basically get that the two maps that should be homotopic are such after post-composition with the inclusion into the total space) $\endgroup$ Jan 26, 2017 at 23:23

1 Answer 1


This is done explicitly by Zabrodsky in his book Hopf Spaces, although I cannot find the exact theorem.

First note that we need a more strict definition of an H-map. An H-map $(f,F):(X,\mu_X)\rightarrow (Y,\mu_Y)$ is a map $f:X\rightarrow Y$ and a homotopy $F:X\times X\rightarrow Y^I$ relative $X\vee X$, satisfying $e_0\circ F=f\circ\mu_x$ and $e_1\circ F=\mu_Y\circ (f\times f)$, where for $a=0,1$, the map $e_a:Y^I\rightarrow Y$ is the evaluation at $a$.

Now the homotopy fibre $F_f$ of $f$ is the (categorical) pullback of $(X\xrightarrow{f} Y\xleftarrow{e_0} PY)$ where $PY=\{l:[0,1]\rightarrow Y\,|\,l(1)=\ast \}$ is the path space over $Y$. That is, $F_f=\{(x,l)\in X\times PY\,|\,f(x)=l(0)\}$. Define $\mu_f:F_f\times F_f\rightarrow F_f$ by


It is straightforward to check that this gives a well defined multiplication on $F_f$. Moreover the canonical projection $F_f\rightarrow X$ and the fibre inclusion $\Omega Y\rightarrow F_f$ are H-maps. The multiplication $\mu_f$ depends on the choice of homotopy $F$. If $\mu_X$, $\mu_Y$ have other properties (homotopy associativity, homotopy commutativity, etc) and $f$ preserves these, then often $\mu_f$ too can be shown to inherit these properties too.

  • $\begingroup$ Thanks for your reply. I am not entirely convinced though, for some reasons. The first being the fact that to even say "F is rel $X \vee X$ " we need to rigidify the unit condition (we know it's always doable by choosing a suitable operation homotopic to the given one), but this would not work in the presence of more structure (say homotopies witnessing associativity). Secondly, the ad-hoc nature of the argument does not (imho) look like the right thing to do. Finally, I managed to get the operation myself, but then how do you prove unitality? $\endgroup$ Jan 27, 2017 at 1:34
  • $\begingroup$ You define an H-space as a unital magma in the based category $ho\mathcal{T}_*$, so the unit should be strict by that definition and $X$, $Y$ should be based by their units by definition. Likewise the H-map $f$ should be based so the relative $X\vee X$ condition on $F$ makes sense. When I say that $F_f$ will inherit any further properties if $f$ preserves them, I really mean that $f$ is given the additional structure of, for example, an $A_n$- or $C_n$-form rather than just the H-map structure $F$. $\endgroup$
    – Tyrone
    Jan 27, 2017 at 17:28

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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