mapping cylinder

Given a continuous map $f:X\to Y$ of topological spaces, one can define its **mapping cylinder** as a pushout

$\array{
X &\stackrel{f}\to& Y
\\
{}^{\mathllap{\sigma_0}}\downarrow && \downarrow^{\mathrlap{ f_*(\sigma_0)}}
\\
X\times I &\stackrel{(\sigma_0)_* (f)}\to & Cyl(f)
}$

in Top, where $I = [0,1]$ (the unit interval) and $\sigma_0:X\to X\times I$ is given by $x\mapsto (x,0)$. By tradition, homotopy theorists sometimes use the inverted (upside-down) mapping cylinder where $\sigma_0$ is replaced by $\sigma_1:x\mapsto (x,1)$. Of course the two mapping cylinders are homeomorphic so it is matter of convention which one to use, of course, compatibly with other constructions depending on the orientation of $I$.

Set-theoretically, the mapping cylinder is usually represented as the quotient space $(X\times I \coprod Y)/{\sim}$ where $\sim$ is the smallest equivalence relation identifying $(x,0)\sim f(x)$ for all $x\in X$.

As any other pushout, the mapping cylinder has a universal property: for any space $Z$ and mapping $g_1:X\times I\to Z$, $g_2:Y\to Z$ such that $g_1(x,0)=g_2(f(x))$ for all $x\in X$, there is a unique $k:Cyl(f)\to Z$, such that the composition $X\times I\to Cyl(f)\stackrel{k}\to Z$ equals $g_1$ and the composition $Y\to Cyl(f)\stackrel{k}\to Z$ equals $g_2$.

Let $f:X\to Y$ be any continuous map. The canonical map $j:=f_*(\sigma_0):Y\to Cyl(f)$ is a homotopy equivalence. In fact its homotopy inverse can be chosen a deformation retraction.

We exhibit $j$ as a homotopy equivalence by constructing its homotopy inverse $\tilde{f}$ given by $\tilde{f}:[x,t]\mapsto f(x)$, where $[x,t]$ is a class of $(x,t)\in X\times I$ and $\tilde{f}([y])=[y]$ for $y\in Y$. Clearly this map is well-defined and $\tilde{f}\circ j = \id_Y$. On the other hand, $(j\circ\tilde{f})[x,t] = [f(x)]$. Homotopy $H:\mathrm{Cyl}(f)\times I\to Y$ is given by

$H([x,t],\tau) = [x,t(1-\tau)], \,\,\,H([y],\tau)=[y].$

It is easy to see that $H(-,0) = \id_{Cyl(f)}$, $H(-,1)=[-,0]=[f(-)]$ hence $j\circ\tilde{f}\sim id_{Cyl(f)}$.

A continuous map $i:A\to X$ is a Hurewicz cofibration iff there is a retraction $r:X\times I\to Cyl(f)$ for the canonical map $X\times I \to Cyl(f)$.

A continuous map $f:X\to Y$ is a homotopy equivalence iff $X = X\times\{0\}$ is a deformation retract of the cylinder $Cyl(f)$.

For any $f:X\to Y$, the composition

$X\stackrel{\sigma_1}\to X\times I\stackrel{(\sigma_0)_* (f)}\to Cyl(f)$

is a Hurewicz cofibration. Furthermore, the map $r:Cyl(f)\to Y$ determined by $r([x,t])= f(x)$ (for all $x\in X$ and $t\in I$) and $r([y])=y$ (for $y\in Y$) is well defined and a homotopy equivalence.

The composition $r\circ (\sigma_0)_* (f)\circ \sigma_1 = f$, hence this is a decomposition of a continuous map into a cofibration followed by a homotopy equivalence.

See also mapping cone.

In homotopy type theory mapping cyclinders can be constructed as higher inductive types. See here.

Revised on February 26, 2015 13:31:02
by Urs Schreiber
(80.92.246.195)