Morphisms from simplicial sets into groupoids are determined by what?

56 Views Asked by At

For this question, one can assume that the simplicial sets come from oriented simplicial complexes (you can assume the orientation induces an orientation on the geometric realization).

A morphism from a simplicial $X$ set into a groupoid $G$ is determined by the restriction of the morphism to the 2-skeleton of $X.$ In addition, a morphism $$f:\Delta^2\to G$$ is determined by $$f:\Lambda_i^2\to G,$$ for any $0\le i\le 2.$ This means that a morphism from the standard 2 simplex to $G$ is determined by the morphism on any horn of the standard 2 simplex, which is a one dimensional simplicial complex. Is something like this true if we replace $\Delta^2$ by any other simplicial complex? That is, given a simplicial complex $X,$ is there a one dimensional subsimplicial complex $Y\subset X$ such that $$\text{Hom}(X,G)=\text{Hom}(Y,G)?$$

In certain examples I can find such a $Y$ by deleting enough edges. My idea was to use the ordering to choose which edges to delete, but I'm not sure if this works in general.

1

There are 1 best solutions below

1
On

There is a universal groupoid that every simplicial set maps to: its fundamental groupoid. To be precise, if $X$ is a simplicial set, consider the groupoid $\pi_{\leq 1}(X)$ whose objects are the vertices of $X$ and whose morphisms are homotopy classes of paths between them in the geometric realization. There is an obvious map $X\to\pi_{\leq 1}(X)$ and every map from $X$ to a groupoid factors uniquely through this map, the edges of $X$ generate $\pi_{\leq 1}(X)$ as a groupoid and homotopic paths formed by sequences of edges become equal under any map to a groupoid.

So, identifying $\operatorname{Hom}(X,G)$ with $\operatorname{Hom}(\pi_{\leq 1}(X),G)$, you are equivalently asking for a $1$-dimensional $Y\subseteq X$ such that $\pi_{\leq 1}(Y)=\pi_{\leq 1}(X)$. Since the fundamental group of a graph is always free, this is impossible unless the fundamental group of $X$ at each basepoint is free.

(I suspect this necessary condition is also sufficient. It would suffice to show that (assuming without loss of generality that $X$ is connected) you can find a $1$-dimensional subcomplex for which the inclusion induces an isomorphism on fundamental groups, since then you can extend that $1$-dimensional subcomplex to include every vertex without changing the fundamental group by adding edges one by one, as in the construction of a spanning tree.)