Let $\pi _1:SS\to Grpd$ denote the fundamental groupoid functor, from simplicial sets to groupoids, and let $N:Grpd\to SS$ denote the nerve functor. Then $\pi _1$ is left adjoint to $N.$

On simplicial sets we consider the usual model category and on groupoids the model category structure inherited by the one given by Thomason. It is known that $\pi _1$ preserves weak equivalences, cofibrations, fibrations and colimits, and that $N$ preserves weak equivalences, fibrations and limits. From these facts one can easily deduce that $\pi _1$ preserves homotopy colimits and $N$ preserves homotopy limits. I wonder whether the following questions have positive answers:

  1. Does $N$ preserve homotopy colimits?
  2. Does $\pi _1$ preserve homotopy limits?
  • $\begingroup$ The nerve functor preserves coproducts. By a result of Dwyer and Kan, the nerve functor preserves certain pushouts. Explicitly, for $G,H,K$ groupoids and for $f:G\to H$ and $g:G\to K$ groupoid homomorphisms, if (1) these groupoids have the same set of objects ($G_0=H_0=K_0$) and (2) both $f_0:G_0\to H_0$ and $g_0:G_0\to K_0$ is the identity morphism on objects, then $N(H\coprod_G K) \simeq NH\coprod_{NG} NK$. $\endgroup$ – user2529 Jul 13 '14 at 15:29
  • $\begingroup$ The fundamental group preserves products. $\endgroup$ – user2529 Jul 13 '14 at 15:29
  • $\begingroup$ Thanks for your answer, @colin. So if I understood well, as a consequence we have that the nerve functor preserves (homotopy) pushouts of groups, right? What is the reference for Dwyer and Kan's result? $\endgroup$ – user55956 Jul 13 '14 at 18:32
  • $\begingroup$ @ColinTan Yes, but not all pushouts are homotopy pushouts. $\endgroup$ – Zhen Lin Jul 13 '14 at 19:39

Just for fun, here's a purely abstract way of seeing neither of these can happen:

The functors you wrote down both preserve weak equivalences, so are morphisms of relative categories, and induce adjunctions of the associated $\left(\infty,1\right)$-categories, which are $\infty$-groupoids and $1$-groupoids (which is actually a $(2,1)$-category) respectively. The adjunction is precisely the one exhibiting $1$-groupoids as a reflective subcategory of $\infty$-groupoids. The full and faithful inclusion is modeled by $N$. The $\left(\infty,1\right)$-category of $\infty$-groupoids is freely generated by the (any) contractible $\infty$-groupoid (i.e. the one point space), (i.e. it's the free colimit cocompletion of the terminal category). Since the $\left(2,1\right)$-category of groupoids is cocomplete, and contains the terminal groupoid, if the inclusion preserved colimits, it would have to be essentially surjective:

If $X$ is any $\infty$-groupoid, you can write $X$ as the colimit of the terminal functor $X \to \infty\mbox{-}\mathbf{Gpd}$ (the constant functor with value the point) (this is basically just the Yoneda Lemma for $\infty$-categories), and this functor (obviously) factors through the inclusion $\mathbf{Gpd} \hookrightarrow \infty\mbox{-}\mathbf{Gpd}$.

If the left adjoint $\tau_1$ of this inclusion (which is modeled by $\Pi_1$) preserved even finite limits, then this would exhibit the $(2,1)$-category of groupoids as a left-exact localization of $\infty\mbox{-}\mathbf{Gpd}=\mathbf{Psh}_\infty\left(*\right),$ i.e. it would exhibit $\mathbf{Gpd}$ as an $\infty$-topos. This means there would exist a unique Grothendieck topology $J$ on the terminal category such that $\mathbf{Gpd}$ sat somewhere between $\mathbf{Sh}_\infty\left(*,J\right)$ and its hypercompletion $\mathbf{HSh}_\infty\left(*,J\right).$ But, there is a unique Grothendieck topology on the terminal category (the trivial one!), so we'd have $$\mathbf{HSh}_\infty\left(*,J\right)=\mathbf{Sh}_\infty\left(*,J\right)=\mathbf{Psh}_\infty\left(*\right).$$ So we'd have to have $\mathbf{Gpd}\simeq \mathbf{Psh}_\infty\left(*\right)=\infty\mbox{-}\mathbf{Gpd}$ which is clearly nonsense.

  • 2
    $\begingroup$ For another concrete example: let $S^1$ be the 1-groupoid which is the fundamental groupoid of the circle and form the homotopy pushout where both maps are $S^1 \to *$. If you form this pushout in 1-groupoids you get something equivalent to the terminal groupoid; if you do it in SS you get $S^2$. $\endgroup$ – Chris Schommer-Pries Jul 14 '14 at 8:36

The fundamental groupoid does not preserve homotopy limits. The simplest examples can be obtained from the path-space sequence $\Omega X\to \operatorname{pt}\to X$ which shows that $\Omega X$ is the homotopy limit of $\operatorname{pt}\to X\leftarrow \operatorname{pt}$. If $\pi_1$ preserves homotopy limits, then the fundamental groupoid of $\Omega X$ should be contractible for any simply-connected space $X$. But any abelian group can be realized as $\pi_2(X)$ for $X$ simply-connected, so the fundamental groupoid can not preserve homotopy limits.

  • $\begingroup$ That's right, thank you! If we restrict to aspherical spaces, would we reach to the same conclusion? $\endgroup$ – user55956 Jul 13 '14 at 18:45
  • $\begingroup$ The fundamental groupoid should preserve homotopy limits for aspherical spaces, I guess for the same reason as in the comment to Zhen Lin's answer. $\endgroup$ – Matthias Wendt Jul 13 '14 at 21:02

The nerve functor does not preserve homotopy colimits. Indeed, take any simplicial set $X$ with non-trivial $\pi_n$ ($n > 1$) and consider $X$ as a simplicial diagram of sets. In $\mathbf{sSet}$, its homotopy colimit will be $X$ (up to weak equivalence, by the Bousfield–Kan theorem), but the nerve of any 1-groupoid has trivial $\pi_n$ ($n > 1$).

The fundamental groupoid functor does not preserve homotopy limits. As Matthias Wendt explained, the loop space can be computed by a homotopy limit, so if we choose a simplicial set $X$ with trivial $\pi_0$ and $\pi_1$ but non-trivial $\pi_2$, then the homotopy pullback diagram $$\begin{array}{ccc} \Omega X & \rightarrow & \Delta^0 \\ \downarrow & & \downarrow \\ \Delta^0 & \rightarrow & X \end{array}$$ cannot be preserved by the fundamental groupoid functor.

  • $\begingroup$ Thank you! And if we restric to aspherical spaces? $\endgroup$ – user55956 Jul 13 '14 at 18:49
  • $\begingroup$ I suppose you mean spaces with trivial $\pi_n$ for $n > 1$? Then yes, because 1-groupoids are a reflective $(\infty, 1)$-subcategory of $\infty$-groupoids. $\endgroup$ – Zhen Lin Jul 13 '14 at 19:44

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.