nLab elegant Reedy category

Definition

Definition

An elegant Reedy category is a Reedy category $R$ such that the following equivalent conditions hold

• For every monomorphism $A\hookrightarrow B$ of presheaves on $R$ and every $x\in R$, the induced map $A_x \amalg_{L_x A} L_x B \to B_x$ is a monomorphism.

• Every span of codegeneracy maps in $R_-$ has an absolute pushout in $R_-$.

• Every element of a presheaf $R$ is a degeneracy of some nondegenerate element in a unique way.

Properties

The principal theorem about elegant Reedy categories is that the Reedy model structure on presheaves (i.e. contravariant diagrams) over an elegant Reedy category coincides with the injective model structure. This is not true for presheaves valued in any model category, only well-behaved ones. We clarify the necessary conditions by building up to this theorem in stages, adding hypotheses on the codomain of the presheaves as necessary.

Degeneracies are split

Lemma

If $R$ is elegant, then every codegeneracy map (i.e. morphism in $R_-$) is a split epimorphism.

Proof

Let $f:x\to y$ be a codegeneracy map; then the span $y \xleftarrow{f} x \xrightarrow{f} y$ has an absolute pushout, consisting of say $g:y\to z$ and $h:y\to z$ with $g f = h f$. This absolute pushout is preserved by $R(z,-)$, so $1_z\in R(z,z)$ must be the image under $g$ or $h$ of some map $s:z\to y$; WLOG say it is $g$, so we have $1_z = g s$. Now we have $s h:y\to y$ and $1_y$ such that $g s h = h = h 1_y$, and our absolute pushout is preserved by $R(y,-)$, so there must be a zigzag of elements in $R(y,z)$ relating $s h$ to $1_y$. At one end of that zigzag, there must be a $t:y\to x$ such that $f t = 1_y$; hence $f$ is split epi.

Corollary

For every monomorphism $A\hookrightarrow B$ of presheaves on $R$, every nondegenerate element of $A$ remains nondegenerate in $B$.

Proof

Let $a$ be a nondegenerate element of $A_x$, for some $x\in R$, $f : x \to y$ a codegeneracy map, and $b\in B_y$ such that $B_f b = a$. We have to show that $f = \id$. By Lemma , $f$ has a section $s: y \to x$, hence $B_s a = B_s B_f b = b$, which implies that $b \in A_y$. Since $a$ is nondegenerate, it follows that $f = \id$.

Degeneracies in subobjects

Lemma

Let $R$ be elegant and $f:x\to y$ a codegeneracy in $R$. Let $M$ be any category, and $\mu:A\to B$ a monomorphism in $M^{R^{\mathrm{op}}}$. Then the following naturality square is a pullback:

$$\array{ A_y & \xrightarrow{A_f} & A_x \\ {}^{\mu_y}\downarrow & & \downarrow^{\mu_x} \\ B_y & \xrightarrow{B_f} & B_x }$$

Proof

This depends only on the fact that $f$ is split epi in $R$. Let $s:y\to x$ be a section of it, and let $P$ be the pullback of $B_f$ and $\mu_x$, with projections $p:P\to A_x$ and $q:P\to B_y$ with $\mu_x p = B_f q$, and an induced map $\phi:A_y \to P$ such that $p \phi = A_f$ and $q\phi = \mu_y$.

We claim that $A_s p : P \to A_y$ is an inverse of $\phi$, making it an isomorphism. On the one hand we have $A_s p \phi = A_s A_f = A_{f s} = 1$. On the other, to show that $\phi A_s p = 1$ it suffices to show that $p \phi A_s p = p$ and $q \phi A_s p = q$. For the first, since $\mu_x$ is monic, it suffices to show $\mu_x p \phi A_s p = \mu_x p$, and for that we have

$$\mu_x p \phi A_s p = B_f q \phi A_s p = B_f \mu_y A_s p = B_f B_s \mu_x p = B_{f s} \mu_x p = \mu_x p.$$

And for the second, we have

$$q \phi A_s p = \mu_y A_s p = B_s \mu_x p = B_s B_f q = B_{f s} q = q.$$

Lemma

Let $R$ be elegant, $M$ a category with pullback-stable colimits, and $\mu:A\to B$ a monomorphism in $M^{R^{\mathrm{op}}}$. Then for any object $x\in R$, the following square is a pullback, where $L_x$ denotes the Reedy latching object at $x$:

$$\array{ L_x A & \to & A_x \\ {}^{L_x \mu}\downarrow & & \downarrow^{\mu_x} \\ L_x B & \to & B_x. }$$

Proof

The map $L_x B \to B_x$ is by definition the colimit in $M/B_x$ of a diagram whose objects are morphisms of the form $B_f : B_y \to B_x$, for $f$ a codegeneracy. By the Lemma , each of these pulls back along $\mu_x$ to $A_f : A_y \to A_x$, forming the corresponding diagram whose colimit is $L_x A \to A_x$, and by assumption the pullback preserves the colimit.

All presheaves are “Reedy monomorphic”

Lemma

Let $R$ be elegant and let $M$ be an infinitary-coherent category. Then for any $x\in R$ and $A\in M^{R^{\mathrm{op}}}$, the map $L_x A \to A_x$ is a monomorphism.

Proof

We use the terminology from the page ∞-ary exact category. Consider the sink with target $A_x$ consisting of all morphisms $A_f : A_y \to A_x$ indexed by nonidentity codegeneracies $f$ with domain $x$. By assumption, for any two such $f:x\to y$ and $f':x\to y'$ there is an absolute pushout $g:y\to z$ and $g':y'\to z$. By absoluteness, $A_z$ is the pullback $A_y \times_{A_x} A_y$. Thus, the images of these absolute pushouts form the kernel of this sink.

Now $L_x A$ is the colimit of the diagram whose objects are $A_y$ indexed by such $f:x\to y$ and whose morphisms are $A_g: A_{y'} \to A_{y}$ for $g:y\to y'$ a codegeneracy with $g f = f'$. In this case, by the universal property of pullback, we have a unique map from $A_{y'}$ to $A_z$, where $z$ is the absolute pushout of $f$ and $f'$. Thus, a cocone under the above kernel is also a cocone under this diagram, and the converse is easy to see. Hence, $L_x A$ is the quotient of the above kernel.

However, in any infinitary-regular category, the quotient of the kernel of a sink is exactly the extremal-epic / monic factorization of that sink. Therefore, the induced map $L_x A \to A_x$ is monic.

Reedy = injective

Theorem

If $R$ is elegant and $M$ is a Grothendieck topos, then for any $x\in R$ and monomorphism $\mu:A\to B$ in $M^{R^{\mathrm{op}}}$, the induced map $L_x B \sqcup_{L_x A} A_x \to B_x$ is monic.

Proof

Since Grothendieck toposes are infinitary-coherent, by Lemma $L_x B\to B_x$ is monic. By assumption $A_x \to B_x$ is monic. And since Grothendieck toposes have pullback-stable colimits, by Lemma the square

$$\array{ L_x A & \to & A_x \\ {}^{L_x \mu}\downarrow & & \downarrow^{\mu_x} \\ L_x B & \to & B_x. }$$

is a pullback. In other words, $L_x A$ is the intersection of the subobjects $L_x B$ and $A_x$ of $B_x$. But in any coherent category, the pushout of two subobjects over their intersection is their union, and hence in particular a subobject of their common codomain.

Corollary

If $R$ is elegant and $M$ is a Cisinski model category, then the Reedy model structure on $M^{R^{\mathrm{op}}}$ coincides with the injective model structure.

Proof

By definition, they have the same weak equivalences, so it suffices to show that their classes of cofibrations coincide. But every Reedy cofibration in any Reedy model structure is an injective (i.e. objectwise) cofibration, and the converse is Theorem .

The most common application is when $M = SSet$. Thus, for instance, every simplicial presheaf on an elegant Reedy category is Reedy cofibrant.

Examples

• The simplex category $\Delta$ is an elegant Reedy category.

• Joyal’s disk categories $\Theta_n$ are elegant Reedy categories.

• Every direct category is a Reedy category with no degeneracies, hence trivially an elegant one.

• If $X$ is any presheaf on an elegant Reedy category $R$, then the opposite of its category of elements $(el X)^{op}$ is again an elegant Reedy category. This is fairly easy to see from the fact that $Set^{el X}$ is equivalent to the slice category $Set^{R^{op}}/X$.

• Every EZ-Reedy category that is a strict Reedy category is elegant.

Note that unlike the notion of Reedy category, the notion of elegant Reedy category is not self-dual: if $R$ is elegant then $R^{op}$ will not generally be elegant.

Related concepts

• generalized Reedy category, Eilenberg-Zilber category

• Reedy model structure

References

• Julie Bergner and Charles Rezk, Reedy categories and the $\Theta$-construction, arXiv:1110.1066

Elegant Reedy categories are useful to model homotopy type theory.

• Michael Shulman, The univalence axiom for elegant Reedy presheaves (arXiv:1307.6248)

• Michael Shulman, Univalence for inverse diagrams and homotopy canonicity, Mathematical Structures in Computer Science, Volume 25, Issue 5 ( From type theory and homotopy theory to Univalent Foundations of Mathematics ) June 2015 (arXiv:1203.3253, doi:/10.1017/S0960129514000565)

• Benno van den Berg and Ieke Moerdijk, W-types in homotopy type theory, PDF