(Sorry for the crossposting, but I'm really interested in this question).

One can define (Paragraph 1.5, page 10) a fibrant-object structure on a suitable cartesian closed category of topological spaces $\bf Top$, called the $\pi_0$-fibrant structure:

  1. A $\pi_0$-equivalence is a map inducing a bijection at the level of $\pi_0$
  2. A $\pi_0$-fibration is a continuous map $p\colon E\to B$ having the RLP with respect to the map $\{0\}\to [0,1]$ including the 0: [I'm not able to reproduce the diagram, the TeX engine seems not to accept the "array" environment]

Every property defining a fibrant structure can be easily shown in the way you see.

Now I'm interested in extending this. The natural definition for a $\pi_n$-equivalence is a map $A\to B$ inducing isomorphisms $\pi_i(A)\to \pi_i(B)$ for all $0\le i\le n$.

What should a $\pi_n$-fibration be in order to define a fibrant structure $\pi_n\text{-}\bf Top$ for all $n\in\mathbb N$?

What if we "go to the limit" (and can it be done?) $\varinjlim_n \big(\pi_n\text{-}\bf Top\big)$ of these fibrant structures? Do we recover a known fibrant structure, obtained forgetting cofibrations and mutual lifting properties of a suitable model structure, on $\bf Top$?


Your question was addressed in the following paper:

Carmen Elvira-Donazar and Luis-Javier Hernandez-Patricio. Closed model categories for the $n$-type of spaces and simplicial sets. Math. Proc. Camb. Phil. Soc. (1995), 118, 93.

Allow me to define an $n$-fibration by quoting from the introduction: Let $I^p$ be the $p$-dimensional unit cube, $V^{p-1}$ be the union of all faces of $I^p$ except for $I^p\times \{1\}$ and $\partial I^p$ the boundary of $I^p$. A map $f$ is an $n$-fibration if it has the right lifting property with respect to $V^{p-1}\to I^p$ (for $0 < p \leq n+1$) and with respect to $V^{n+1}\to \partial I^{n+2}$.

With this definition, and your notion of an $n$-equivalence they prove that $Top$ (meaning a suitable cartesian-closed version) is a model category. So you can forget all mention of cofibrations and get the fibrant-object structure you wanted. The proof proceeds by way of Simplicial Sets, so if you read that paper you'll probably learn loads more about $n$-fibrations. For instance, Corollary 2.1 says trivial $n$-fibrations are exactly maps which have the RLP with respect to $\partial I^p\to I^p$ for $0\leq p\leq n+1$.

It is not difficult to see from the description of $n$-equivalences and $n$-fibrations that in the limit as $n\to \infty$ you get the usual model structure on $Top$. I should mention that this paper of Golasinski and Goncalves credits this model structure to Tim Porter and J.L. Hernandez via Categorical models of $n$-types for procrossed complexes and $\mathcal{J}_n$-prospaces from the 1990 Barcelona Conference on Algebraic Topology. But I couldn't find an online copy of that, so I went with the reference above instead.

Note that the dual question to your question (declaring $X\to Y$ to be an $n$-equivalence if $\pi_k(X)\to \pi_k(Y)$ is an isomorphism for all $k>n$) has also been answered, and again there is a model structure. Here is a reference:

J. Ignacio Extremiana Aldana, L. Javier Hernández Paricio, and M. Teresa Rivas Rodríguez. A closed model category for ($n-1$)-connected spaces. Proc. Amer. Math. Soc. 124 (1996), 3545-3553

EDIT (April 1, 2013):

I recently learned of another paper in this vein by the same authors, thanks to a comment of Fernando Muro over at this MO thread. Here is the reference:

J. Ignacio Extremiana Aldana , L. Javier Hernández Paricio , M. Teresa Rivas Rodríguez. Closed Model Categories For [n,m]-Types (1997)

This combines the two types of truncation mentioned above to get a model structure on [n,m]-types (truncated by $n$ below and $m$ above) whose homotopy category is equivalent to the category of $n$-reduced CW complexes with dimension $\leq m+1$ and $m$-homotopy classes of maps. It actually does even more, because it gives a different model structure whose homotopy category is equivalent to the homotopy category of $(n-1)$-connected, $(m+1)$-coconnected CW complexes.

  • $\begingroup$ maybe you mean "$V^{p-1}$ = the union of all faces of $I^p$ except for $I^{p-1}\times 1$"? $\endgroup$ – Fosco Dec 17 '12 at 19:51
  • $\begingroup$ ...I'm totally uncomfortable with the definition of $V^{p-1}$. Can you write explicitly $V^{-1}$ (if it can be defined), $V^0,V^1, V^2$? $\endgroup$ – Fosco Dec 17 '12 at 20:01
  • $\begingroup$ Did you look at the linked paper? Maybe I made an error transcribing that description. The paper goes into more detail on this construction. Perhaps it's supposed to be "except for $I^{p-1}\times 1$." $\endgroup$ – David White Dec 17 '12 at 20:17
  • $\begingroup$ It's obviously $I^{p-1}$; I checked and the error is repeated verbatim in the version of the paper you linked me. $\endgroup$ – Fosco Dec 18 '12 at 22:26
  • $\begingroup$ (which is the submitted version, but it's strange, isn't it?) $\endgroup$ – Fosco Dec 18 '12 at 22:27

There was a mistake in an earlier version of the paper that you mention. If you define $\pi_0$-fibrations and $\pi_0$-equivalences the way that you did, they do not give the structure of a category of fibrant objects on $\bf Top$. The reason is that then the acyclic fibrations will not be closed under pullbacks, as they should be by (FW1) loc. cit.

A counterexample is the map $R\to S^1$ of the real line to the circle that gives the universal cover ($t\to \exp(2\pi it)$). This is a $\pi_0$-acyclic fibration. Take the pullback of this map with respect to $*\to S^1$. The pullback is the fiber of the first map at the point, which is $Z$ (integers). The map from $Z$ to the point $*$ is not a $\pi_0$-equivalence.

I remarked this to Otgonbayar Uuye and he fixed this in the new version. The main conclusion he wanted to draw from it, namely, that Schochet fibrations and homotopy equivalences form a category of fibrant objects on $C^*$-algebras, still holds. The point is to use the standard (Quillen) model structure on $\bf Top$.

As for larger $n$, we can begin with the Quillen model structure on $\bf Top$, and take its left Bousfield Localization with respect to the map $S^{ n+1}\to *$. We obtain a new model structure on $\bf Top$ (so restricting to the fibrant objects we get a fibrant objects structure), with the weak equivalences being the $\pi_n$-equivalences.

Note, however, that there are less fibrations in this model structure then in the usual one. In other words, any $\pi_n$-fibration would also be a Serre fibration. For e.g., the fibrant objects in the localized model structure would be (not all spaces, but only) spaces $X$ such that $\pi_i(X)=0$ for $i> n$.

  • $\begingroup$ $R\to S^1$ is an acyclic $\pi_0$-fibration? $\endgroup$ – David Roberts Aug 4 '15 at 7:00
  • $\begingroup$ It induces isomorphism on $\pi_0$ so it is a $\pi_0$ equivalence. It is clearly also a $\pi_0$-fibration. $\endgroup$ – Ilan Barnea Aug 4 '15 at 7:08
  • $\begingroup$ I meant: it's not an acyclic fibration in the usual sense! $\endgroup$ – David Roberts Aug 4 '15 at 7:38
  • $\begingroup$ Yes, of course. Only in the fibrant object structure defined in the question $\endgroup$ – Ilan Barnea Aug 4 '15 at 11:08

This might be a naive answer but here is a suggestion for the definition of $\pi_n$-fibrations: maps having the RLP with respect to the the map $\Delta^k\to\Delta^k\times I$ for any $k\leq n$.

In the limit you will get the "obvious" fibrant-object structure on $Top$ which comes from its usual model structure (recall that the full subcategory of fibrant objects in any model category is a category of fibrant objects.. and that all objects are fibrants in $Top$).


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.