Let $B \leftarrow A \to C$ be a span of spaces, and consider the homotopy pushout $B \cup_A C$.

Question: When is $B \cup_A C$ contractible?

This is a pretty open-ended question. I'm interested in necessary conditions or sufficient conditions or interesting examples or special cases, etc.

Some additional conditions I'd be happy to take as blanket assumptions:

  1. I don't think too much is lost if we assume that $A,B,C$ are connected.

  2. I'm happy to assume that $A \to B$ is 1-connected (conventions vary about what this means; I mean that the homotopy fiber of $A \to B$ is 1-connected, i.e. that $A \to B$ is $\pi_1$-surjective).

  3. For somewhat obscure reasons, I'm particularly interested in the case where $A \to C$ is the projection $A_0 \times C \to C$ out of a binary product.

  4. For similarly obscure reasons, I'm particularly interested in the case where $C$ is a loopspace, or even where $C = \Omega \Sigma C_0$ is a free loopspace.


  • By pasting on a few more pushout squares, we obtain the identity $\Sigma(B \vee C) = \Sigma A$.

I'm not sure what to make of this, though.

  • 1
    $\begingroup$ You've probably already thought of this, but you can get some algebraic necessary conditions by using Mayer-Vietoris in your favorite (co)homology theory. $\endgroup$ – John Greenwood Jan 22 at 2:56
  • 1
    $\begingroup$ @John I think this is covered already by Tim's observation, which is an almost unstable version of what MV gives you $\endgroup$ – Phil Tosteson Jan 22 at 4:07
  • 1
    $\begingroup$ @Tim Knots are one rich source of examples. If $K \subset S^3$ is a knot there is a pushout square with $C = S^3 - K$, $B$ a tubular neighborhood of the knot in $\mathbb R^3$ (contractible), and $A$ is $B - K$ (equivalent to $S^1$). However, this doesn't produce many examples which satisfy $4$, because $\pi_1 C$ is rarely abelian. $\endgroup$ – Phil Tosteson Jan 22 at 4:23
  • 5
    $\begingroup$ The paper Structure theorems for homotopy pushouts. I. Contractible pushouts. by John Klein (MR1490203) is all about contractible pushouts. $\endgroup$ – Jeff Strom Jan 22 at 17:08

Assume all the spaces are connected, $\pi_1 A \to \pi_1 B$ is surjective, and $\pi_1 C$ is abelian.

The pushout, $P = C \sqcup_A B$, is contractible if and only if $\pi_1(P)$ and all the $H_i(P)$ are trivial. However, by Seifert-Van-Kampen, and the hypothesis on $\pi_1 A \to \pi_1 B$, we have that $\pi_1 C$ surjects onto $\pi_1 P$. Thus $\pi_1 P$ is abelian, so $P$ is contractible if and only $H_i(P) = 0$ for all $i$.

For this, as you and John observed, it is necessary and sufficent for $H_i(A) \to H_i(B) \oplus H_i(C)$ to be an isomorphism for all $i \geq 1$.


Let me add some additional remarks on the enumeration question:

How many such spaces $A$ are there sitting over $B\times C$ such that the homotopy pushout $$ B \leftarrow A \to C $$ is contractible?

  1. If $B\cup_A C$ is contractible, then the sum of the maps $\Sigma A \to \Sigma B$ and $\Sigma A \to \Sigma C$ gives a homology isomorphism $$ \Sigma A \to \Sigma B \vee \Sigma C\, . $$ If $A$ is connected, then we conclude that the map is also a weak homotopy equivalence, so $\Sigma A$ is a wedge of $\Sigma B$ and $\Sigma C$ in this case. The converse to this statement is also true.

  2. Hence, we are reduced to classifying those spaces $A$ whose suspension splits as a wedge of $\Sigma B$ and $\Sigma C$. Hence, this problem can be formulated as a kind desuspension problem.

  3. There is a classifying space ${\cal D}(B,C)$ which is the realization of (the nerve of) a category whose objects are spaces $A$ with structure map $A \to B\times C$ such that $\text{hocolim}(B \leftarrow A \to C)$ is contractible. A morphism of this category is a weak homotopy equivalence of spaces over $B\times C$.

  4. Classification Result: Assume $A,B,C$ are $1$-connected. There is a function $$ \pi_0({\cal D}(B,C)) \to \{B\vee C, B\wedge C\} $$ (the target is the abelian group of homotopy classes of stable maps) which is a bijection in a metastable range $$ \max(b,c) \le 3\min(r,s)-1\, , $$ where $b$ is the homotopy dimension of $B$ (the CW complex of minimal dimension having the homotopy type of $B$), $c$ is the homotopy dimension of $C$, $r$ is the connectivity of $B$ and $s$ is the connectivity of $C$. Here I have given $B$ and $C$ basepoints.

  5. The function in (4) is easily described as the stable homotopy class of the weak map $$ \Sigma B\vee \Sigma C \overset{\simeq}\leftarrow \Sigma A \to \Sigma (B \wedge C) $$ where the right arrow is given by suspending the composition $A \to B\times C \to B\wedge C$.

  6. If we want to work beyond the metastable range, the homotopy type of ${\cal D}(B,C)$ can be determined "up to extensions" via the coefficients of the identity functor in homotopy functor calculus.

See the paper

Klein, John R.; Peter, John W. Fake wedges. Trans. Amer. Math. Soc. 366 (2014), no. 7, 3771–3786

which extends my earlier paper:

John R. Klein, Structure theorems for homotopy pushouts. I. Contractible pushouts, Math. Proc. Cambridge Philos. Soc. 123 (1998), no. 2, 301–324.


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.