nLab (infinity,1)-pullback

Context

$\left(\infty ,1\right)$-Category theory

(∞,1)-category theory

Contents

Idea

An $\left(\infty ,1\right)$-pullback is a limit in an (∞,1)-category $𝒞$ over a diagram of the shape

$\left\{a\to c←b\right\}\to 𝒞\phantom{\rule{thinmathspace}{0ex}}.$\{a \to c \leftarrow b\} \to \mathcal{C} \,.

In other words it is a cone

$\begin{array}{ccc}A{×}_{C}B& \to & B\\ ↓& \cong ⇙& ↓\\ A& \to & C\end{array}$\array{ A \times_C B &\to& B \\ \downarrow &\cong\swArrow& \downarrow \\ A &\to& C }

which is universal among all such cones in the $\left(\infty ,1\right)$-categorical sense.

This is the analog in (∞,1)-category theory of the notion of pullback in category theory.

Incarnations

In quasi-categories

Let $𝒞$ be a quasi-category. Recall the notion of limit in a quasi-category.

The non-degenerate cells of the simplicial set $\Delta \left[1\right]×\Delta \left[1\right]$ obtained as the cartesian product of the simplicial 1-simplex with itself look like

$\begin{array}{ccc}\left(0,0\right)& \to & \left(1,0\right)\\ ↓& ↘& ↓\\ \left(0,1\right)& \to & \left(1,1\right)\end{array}$\array{ (0,0) &\to& (1,0) \\ \downarrow &\searrow& \downarrow \\ (0,1) &\to& (1,1) }

A square in a quasi-category $C$ is an image of this in $C$, i.e. a morphism

$s:\Delta \left[1\right]×\Delta \left[1\right]\to C\phantom{\rule{thinmathspace}{0ex}}.$s : \Delta[1] \times \Delta[1] \to C \,.

The simplicial square $\Delta \left[1{\right]}^{×2}$ is isomorphic, as a simplicial set, to the join of simplicial sets of a 2-horn with the point:

$\Delta \left[1\right]×\Delta \left[1\right]\simeq \left\{v\right\}\star \Lambda \left[2{\right]}_{2}=\left(\begin{array}{ccc}v& \to & 1\\ ↓& ↘& ↓\\ 0& \to & 2\end{array}\right)$\Delta[1] \times \Delta[1] \simeq \{v\} \star \Lambda[2]_2 = \left( \array{ v &\to& 1 \\ \downarrow &\searrow& \downarrow \\ 0 &\to& 2 } \right)

and

$\Delta \left[1\right]×\Delta \left[1\right]\simeq \Lambda \left[2{\right]}_{0}\star \left\{v\right\}=\left(\begin{array}{ccc}0& \to & 1\\ ↓& ↘& ↓\\ 2& \to & v\end{array}\right)\phantom{\rule{thinmathspace}{0ex}}.$\Delta[1] \times \Delta[1] \simeq \Lambda[2]_0 \star \{v\} = \left( \array{ 0&\to& 1 \\ \downarrow &\searrow& \downarrow \\ 2 &\to& v } \right) \,.

If a square $\Delta \left[1\right]×\Delta \left[1\right]\simeq \Lambda \left[2{\right]}_{0}\star \left\{v\right\}\to C$ exhibits $\left\{v\right\}\to C$ as a quasi-categorical limit over $F:\Lambda \left[2{\right]}_{0}\to C$, we say the limit

$v:=\underset{←}{\mathrm{lim}}F:=F\left(1\right)\prod _{F\left(0\right)}F\left(2\right)$v := \lim_\leftarrow F := F(1) \prod_{F(0)} F(2)

is the quasi-categorical pullback of the diagram $F$.

Pasting law

We have the following quasi-categorical analog of the familiar pasting law of pullbacks in ordinary category theory:

A pasting diagram of two squares is a morphism

$\sigma :\Delta \left[2\right]×\Delta \left[1\right]\to 𝒸\phantom{\rule{thinmathspace}{0ex}}.$\sigma : \Delta[2] \times \Delta[1] \to \mathcal{c} \,.

Schematically this looks like

$\begin{array}{ccccc}a& \to & b& \to & c\\ ↓& & ↓& & ↓\\ d& \to & e& \to & f\end{array}$\array{ a &\to& b &\to& c \\ \downarrow && \downarrow && \downarrow \\ d &\to& e &\to& f }

in $𝒞$.

Proposition

(pasting law for quasi-categorical pullbacks)

If the right square is a pullback diagram in $𝒞$, then the left square is precisely if the outer square is.

This is HTT, lemma 4.4.2.1

Proof

Consider the diagram inclusions

$\left(\begin{array}{ccccc}& & & & c\\ & & & & ↓\\ d& \to & e& \to & f\end{array}\right)\phantom{\rule{thickmathspace}{0ex}}\phantom{\rule{thickmathspace}{0ex}}\to \phantom{\rule{thickmathspace}{0ex}}\phantom{\rule{thickmathspace}{0ex}}\left(\begin{array}{ccccc}& & b& \to & c\\ & & ↓& & ↓\\ d& \to & e& \to & f\end{array}\right)\phantom{\rule{thickmathspace}{0ex}}\phantom{\rule{thickmathspace}{0ex}}←\phantom{\rule{thickmathspace}{0ex}}\phantom{\rule{thickmathspace}{0ex}}\left(\begin{array}{ccc}& & b\\ & & ↓\\ d& \to & e\end{array}\right)$\left( \array{ && && c \\ && && \downarrow \\ d &\to& e &\to& f } \right) \;\;\to\;\; \left( \array{ && b &\to& c \\ && \downarrow && \downarrow \\ d &\to& e &\to& f } \right) \;\;\leftarrow\;\; \left( \array{ && b \\ && \downarrow \\ d &\to& e } \right)

and the induced diagram of over quasi-categories

${𝒞}_{/\sigma \left(c,d,f\right)}\stackrel{\varphi }{←}{𝒞}_{/\sigma \left(b,c,d,e,f\right)}\stackrel{\psi }{\to }{𝒞}_{/\sigma \left(b,d,e\right)}\phantom{\rule{thinmathspace}{0ex}}.$\mathcal{C}_{/\sigma(c,d,f)} \stackrel{\phi}{\leftarrow} \mathcal{C}_{/\sigma(b,c,d,e,f)} \stackrel{\psi}{\to} \mathcal{C}_{/\sigma(b,d,e)} \,.

Notice that by definition of limit in a quasi-category the quasi-categorical pullback $\sigma \left(c\right){×}_{\sigma \left(f\right)}\sigma \left(d\right)$ is the terminal object in ${𝒞}_{/\sigma \left(c,d,f\right)}$, while $\sigma \left(d\right){×}_{\sigma \left(e\right)}\sigma \left(b\right)$ is the terminal object in ${𝒞}_{/\sigma \left(b,d,e\right)}$.

The strategy now is to show that both these morphisms $\varphi$ and $\psi$ are acyclic Kan fibrations. That will imply that these terminal objects coincide as objects of $𝒞$.

First notice that the inclusion

$\left(\begin{array}{ccc}& & b\\ & & ↓\\ d& \to & e\end{array}\right)\phantom{\rule{thickmathspace}{0ex}}\phantom{\rule{thickmathspace}{0ex}}\to \phantom{\rule{thickmathspace}{0ex}}\phantom{\rule{thickmathspace}{0ex}}\left(\begin{array}{ccccc}& & b& \to & c\\ & & ↓& & ↓\\ d& \to & e& \to & f\end{array}\right)$\left( \array{ && b \\ && \downarrow \\ d &\to& e } \right) \;\; \to \;\; \left( \array{ && b &\to& c \\ && \downarrow && \downarrow \\ d &\to& e &\to& f } \right)

is a left anodyne morphism, being the composite of pushouts of left horn inclusions

$\begin{array}{rl}\left(\begin{array}{ccc}& & b\\ & & ↓\\ d& \to & e\end{array}\right)& \to \left(\begin{array}{ccccc}& & b& \to & c\\ & & ↓& & \\ d& \to & e\end{array}\right)\\ & \to \left(\begin{array}{ccccc}& & b& \to & c\\ & & ↓& & ↓\\ d& \to & e& & f\end{array}\right)\\ & \to \left(\begin{array}{ccccc}& & b& \to & c\\ & & ↓& ↘& ↓\\ d& \to & e& & f\end{array}\right)\\ & \to \left(\begin{array}{ccccc}& & b& \to & c\\ & & ↓& ↘& ↓\\ d& \to & e& \to & f\end{array}\right)\end{array}\phantom{\rule{thinmathspace}{0ex}}.$\begin{aligned} \left( \array{ && b \\ && \downarrow \\ d &\to& e } \right) & \to \left( \array{ && b &\to& c \\ && \downarrow && \\ d &\to& e } \right) \\ & \to \left( \array{ && b &\to& c \\ && \downarrow && \downarrow \\ d &\to& e && f } \right) \\ & \to \left( \array{ && b &\to& c \\ && \downarrow &\searrow& \downarrow \\ d &\to& e && f } \right) \\ & \to \left( \array{ && b &\to& c \\ && \downarrow &\searrow& \downarrow \\ d &\to& e &\to& f } \right) \end{aligned} \,.

We could also prove this by showing that this functor is homotopy initial using the characterization in terms of slice categories, and then invoking the theorem of HTT 4.1.1.3(4) which says (in dual form) that an inclusion of simplicial sets is homotopy initial if and only if it is left anodyne.

One of the properties of left anodyne morphisms is that restriction of over quasi-categories along left anodyne morphisms produces an acyclic Kan fibration. This shows the desired statement for $\psi$.

To see that $\varphi$ is also an acyclic fibration observe that $\varphi$ can be factored as

${𝒞}_{/\sigma \left(c,d,f\right)}←{𝒞}_{/\sigma \left(c,d,e,f\right)}←{𝒞}_{/\sigma \left(b,c,d,e,f\right)}$\mathcal{C}_{/\sigma(c,d,f)} \leftarrow \mathcal{C}_{/\sigma(c,d,e,f)} \leftarrow \mathcal{C}_{/\sigma(b,c,d,e,f)}

Observe that ${𝒞}_{/\sigma \left(c,d,e,f\right)}←{𝒞}_{/\sigma \left(b,c,d,e,f\right)}$ fits into a pullback diagram

$\begin{array}{ccc}{𝒞}_{/\sigma \left(c,d,e,f\right)}& ←& {𝒞}_{/\sigma \left(b,c,d,e,f\right)}\\ ↓& & ↓\\ {𝒞}_{/\sigma \left(c,e,f\right)}& ←& {𝒞}_{/\sigma \left(b,c,e,f\right)}\end{array}$\array{ \mathcal{C}_{/\sigma(c,d,e,f)} & \leftarrow & \mathcal{C}_{/\sigma(b,c,d,e,f)} \\ \downarrow & & \downarrow \\ \mathcal{C}_{/\sigma(c,e,f)} & \leftarrow & \mathcal{C}_{/\sigma(b,c,e,f)} }

and hence is an acyclic Kan fibration since ${𝒞}_{/\sigma \left(c,e,f\right)}←{𝒞}_{/\sigma \left(b,c,e,f\right)}$ is one, on account of the fact that the square

$\begin{array}{ccc}\sigma \left(b\right)& \to & \sigma \left(c\right)\\ ↓& & ↓\\ \sigma \left(e\right)& \to & \sigma \left(f\right)\end{array}$\array{ \sigma(b) & \to & \sigma(c) \\ \downarrow & & \downarrow \\ \sigma(e) & \to & \sigma(f) }

is a pullback in $𝒞$. Finally, ${𝒞}_{/\sigma \left(c,d,f\right)}←{𝒞}_{/\sigma \left(c,d,e,f\right)}$ is a trivial fibration since

$\begin{array}{ccc}\left(\begin{array}{ccc}& & c\\ & & ↓\\ d& \to & f\end{array}\right)& \to & \left(\begin{array}{ccccc}& & & & c\\ & & & & ↓\\ d& \to & e& \to & f\end{array}\right)\end{array}$\array{ \left( \array{ & & c \\ & & \downarrow \\ d & \to & f } \right) & \to & \left( \array{ & & & & c \\ & & & & \downarrow \\ d & \to & e & \to & f } \right) }

is left anodyne; clearly this is a pushout of $\left(d\to f\right)\to \left(d\to e\to f\right)$ and so it suffices to show that ${\Delta }^{\left\{0,2\right\}}\to {\Delta }^{\left\{0,1,2\right\}}$ is left anodyne. But this map factors as ${\Delta }^{\left\{0,2\right\}}\to {\Lambda }_{0}^{2}\to {\Delta }^{\left\{0,1,2\right\}}$ and clearly ${\Delta }^{\left\{0,2\right\}}\to {\Lambda }_{0}^{2}$ is left anodyne since it is a pushout of ${\Delta }^{\left\{0\right\}}\to {\Delta }^{\left\{0,1\right\}}$.

Examples

Fiber sequence

If $𝒞$ has a terminal object and $*\to C\in 𝒞$ is a pointed object, then the fiber or $\left(\infty ,1\right)$-kernel of a morphisms $f:B\to C$ is the $\left(\infty ,1\right)$-pullback

$\begin{array}{ccc}\mathrm{ker}\left(f\right)& \to & *\\ ↓& & ↓\\ B& \stackrel{f}{\to }& C\end{array}\phantom{\rule{thinmathspace}{0ex}}.$\array{ ker(f) &\to& * \\ \downarrow && \downarrow \\ B &\stackrel{f}{\to}& C } \,.

For more on this see fiber sequence.

References

In homotopy type theory

A formalization of homotopy pullbacks in homotopy type theory is Coq-coded in

Revised on July 9, 2012 23:11:53 by Danny Stevenson (81.110.14.49)