on chain complexes/model structure on cosimplicial abelian groups
related by the Dold-Kan correspondence
on algebras over an operad, on modules over an algebra over an operad
on dendroidal sets, for dendroidal complete Segal spaces, for dendroidal Cartesian fibrations
A simplicial homotopy is a homotopy in the classical model structure on simplicial sets. It can also be defined combinatorially; in that form one can define a homotopy 2-cell between morphisms of simplicial objects in any category $C$.
SSet has a cylinder functor given by cartesian product with the standard 1-simplex, $I := \Delta[1]$. (In fact, one can define simplicial cylinders, $\Delta[1]\odot X$, more generally, for example for $X$ being a simplicial object in an cocomplete category $C$,(see below).)
Therefore for $f,g : X \to Y$ two morphisms of simplicial sets, a homotopy $\eta : f \Rightarrow g$ is a morphism $\eta : X \times \Delta[1] \to Y$ such that the diagram
commutes.
Remark: Since in the standard model structure on simplicial sets every simplicial set is cofibrant, this indeed defines left homotopies.
Given morphisms $f,g,:X\to Y$ of simplicial objects in any category $C$, a simplicial homotopy is a family of morphisms, $h_i:X_n\to Y_{n+1}$, $i= 0,\ldots,n$ of $C$, such that $d_0 h_0 = f_n$, $d_{n+1} h_n = g_n$ and
The above formulae give one of the, at least, two forms of the combinatorial specfication of a homotopy between $f$ and $g$. (When trying to construct a specific homotopy using a combinatorial form, check which convention is being used!) The two forms correspond to different conventions such as saying that this is a homotopy from $g$ to $f$, or reversing the labelling of the $h_i$.
It is fairly easy to prove that the combinatorial definition of homotopy agrees with the one via the cylinder both for simplicial sets and for simplicial objects in any finitely cocomplete category, $C$. This uses the fact that the category of simplicial objects in a cocomplete category, $C$, has copowers with finite simplicial sets and hence in particular with $\Delta[1]$. (As there are explicit formulae for the construction of copowers …)
Tim: With only my own resources available, I was unable to find them so was hoping someone kind would come up with them. They derive from the coend formulae/Kan extension formulae using some combinatorics to discuss the indexing sets. I think Quillen gave some form of them, but have not got a copy of HA. I needed them recently and could not find them in any of the usual sources, and did not manage to work them out using the Kan extension idea either (Help please anyone). We could do with those formulae or with a reference to them at least.
In the case of the category of (not necessarily abelian) groups, the combinatorial definition equals the one via cylinder only if the role of “cylinder” for a group $G$ is played by a simplicial object in the category of groups which in degree $n$ equals the free product of $(n+2)$ copies of $G$, indexed by the set $\Delta[1]_n$ (noted by Swan and quoted in exercise 8.3.5 of Weibel: Homological algebra).
Tim Porter: Perhaps we need an explicit description of copowers in simplicial objects also. I pointed out in an edit above that the combinatorial description is much more general than just for simplicial objects in an abelian category.
Can specific references to Swan be given, anyone?
Zoran Škoda: I agree that one should talk about copowers etc.
Precisely when $Y$ is a Kan complex, the relation
is an equivalence relation.
Since Kan complexes are precisely the fibrant objects with respect to the standard model structure on simplicial sets this follows from general statements about homotopy in model categories.
The following is a direct proof.
We first show that the homotopy between points $x,y : \Delta[0] \to Y$ is an equivalence relation when $Y$ is a Kan complex.
We identify in the following $x$ and $y$ with vertices in the image of these maps.
-reflexivity- For every vertex $x \in Y_0$, the degenerate 1-simplex $s_0 x \in S_1$ has, by the simplicial identities, 0-faces $d_0 s_0 x = x$ and $d_1 s_0 x = x$.
Therefore the morphism $\eta : \Delta[0] \times \Delta[1] \to Y$ that takes the unique non-degenerate 1-simplex in $\Delta[1]$ to $s_0 x$ constitutes a homotopy from $x$ to itself.
-transitivity- let $v_2 : x \to y$ and $v_0 : y \to z$ in $Y_1$ be 1-cells. Together they determine a map from the horn $\Lambda^2_1$ to $Y$,
By the Kan complex property there is an extension $\theta$ of this morphism through the 2-simplex $Delta^2$
If we again identify $\theta$ with its image (the image of its unique non-degenerate 2-cell) in $Y_2$, then using the simplicial identities we find
that the 1-cell boundary bit $d_1 \theta$ in turn has 0-cell boundaries
and
This means that $d_1 \theta$ is a homotopy $x \to z$.
-symmetry- In a similar manner, suppose that $v_2 : x \to y$ is a 1-cell in $Y_1$ that constitutes a homotopy from $x$ to $y$. Let $v_1 := s_0 x$ be the degenerate 1-cell on $x$. Since $d_1 v_1 = d_1 v_2$ together they define a map $\Lambda^2_0 \stackrel{v_1, v_2}{\to} Y$ which by the Kan property of $Y$ we may extend to a map $\theta'$
on the full 2-simplex.
Now the 1-cell boundary $d_0 \theta'$ has, using the simplicial identities, 0-cell boundaries
and
and hence yields a homotopy $y \to x$. So being homotopic is a symmetric relation on vertices in a Kan complex.
Finally we use the fact that SSet is a cartesian closed category to deduce from this statements about vertices the corresponding statement for all map:
a morphism $f : X \to Y$ is the Hom-adjunct of a morphism $\bar f : \Delta[0] \to [X,Y]$, and a homotopy $\eta : X \times \Delta[1] \to Y$ is the adjunct of a morphism $\bar \eta : \Delta[1] \to [X,Y]$. Therefore homotopies $\eta : f \Rightarrow g$ are in bijection with homotopies $\bar \eta : \bar f \to \bar g$.
Let $A$ be an abelian category and $f,g : X\to Y$ homotopic morphisms of simplicial objects in $A$. Then the induced maps $f_*, g_* : N(X)\to N(Y)$ of their normalized chain complexes are chain homotopic.
Let $A$ be an abelian category. The morphisms of simplicial objects (variant: of unbounded chain complexes) in $A$, which are homotopic to zero, form an ideal. More precisely
being homotopic is an equivalence relation on the class of morphism,
$f_1\sim 0$ and $f_2\sim 0$ implies $f_1+f_2 \sim 0$,
if $f\circ h$ (resp. $h\circ f$) exists and if $f\sim 0$ then $f\circ h\sim 0$ (resp. $h\circ f\sim 0$).
Cylinder based homotopy is also discussed extensively in