Kerodon

$\Newextarrow{\xRightarrow}{5,5}{0x21D2}$ $\newcommand\empty{}$
$\Newextarrow{\xhookrightarrow}{10,10}{0x21AA}$

Theorem 5.4.4.1 (Local Thinness Criterion). Let $\operatorname{\mathcal{C}}$ be an $(\infty ,2)$-category and let $\sigma $ be a $2$-simplex of $\operatorname{\mathcal{C}}$, which we represent by the diagram

\[ \xymatrix@R =50pt@C=50pt{ & Y \ar [dr]^{v} & \\ X \ar [ur]^{u} \ar [rr]^-{w} & & Z. } \]

The following conditions are equivalent:

$(1)$

The $2$-simplex $\sigma $ is thin.

$(2)$

Let $q: \operatorname{\mathcal{C}}_{/Z} \rightarrow \operatorname{\mathcal{C}}$ denote the projection map. Then $\sigma $ is $q$-cartesian when viewed as an edge of the simplicial set $\operatorname{\mathcal{C}}_{/Z}$.

$(3)$

The $2$-simplex $\sigma $ is locally $q$-cartesian when viewed as an edge of the simplicial set $\operatorname{\mathcal{C}}_{/Z}$.

$(4)$

Let $q': \operatorname{\mathcal{C}}_{X/} \rightarrow \operatorname{\mathcal{C}}$ denote the projection map. Then $\sigma $ is $q'$-cocartesian when viewed as an edge of the simplicial set $\operatorname{\mathcal{C}}_{X/}$.

$(5)$

The $2$-simplex $\sigma $ is locally $q'$-cocartesian when viewed as an edge of the simplicial set $\operatorname{\mathcal{C}}_{X/}$.

Proof. We will prove that $(1) \Leftrightarrow (2) \Leftrightarrow (3)$; the proof that $(1) \Leftrightarrow (4) \Leftrightarrow (5)$ follows by applying the same argument to the opposite $(\infty ,2)$-category $\operatorname{\mathcal{C}}^{\operatorname{op}}$. The implication $(1) \Rightarrow (2)$ follows from Proposition 5.4.3.8, and the implication $(2) \Rightarrow (3)$ is immediate (see Remark 5.1.3.3). For each integer $n \geq 3$, consider the following weaker version of condition $(1)$:

$(1_ n)$

For every integer $0 < i < n$ and every morphism of simplicial sets $\mu _0: \Lambda ^{n}_{i} \rightarrow \operatorname{\mathcal{C}}$ for which the composition

\[ \Delta ^{2} \simeq \operatorname{N}_{\bullet }( \{ i-1 < i < i+1\} ) \hookrightarrow \Lambda ^{n}_{i} \xrightarrow {\mu _0} \operatorname{\mathcal{C}} \]

is equal to $\sigma $, there exists a map $\mu : \Delta ^ n \rightarrow \operatorname{\mathcal{C}}$ extending $\mu _0$.

Note that $\sigma $ satisfies condition $(1)$ if and only if it satisfies condition $(1_ n)$ for each $n \geq 3$. We will complete the proof by showing that $(3) \Rightarrow (1_ n)$, using a fairly elaborate induction on $n$.

Assume that $\sigma $ is locally $q$-cartesian when viewed as a morphism in the $(\infty ,2)$-category $\operatorname{\mathcal{C}}_{/Z}$. Since $\operatorname{\mathcal{C}}$ is an $(\infty ,2)$-category, we can choose a thin $2$-simplex $\sigma '$ satisfying $d^{2}_0(\sigma ') = d^{2}_0(\sigma )$ and $d^{2}_2(\sigma ') = d^{2}_2(\sigma )$, which we represent as a diagram

\[ \xymatrix@R =50pt@C=50pt{ & Y \ar [dr]^{v} & \\ X \ar [ur]^{u} \ar [rr]^-{w'} & & Z. } \]

The implication $(1) \Rightarrow (3)$ shows that $\sigma '$ is also locally $q$-cartesian when viewed as an edge of the simplicial set $\operatorname{\mathcal{C}}_{/Z}$. Let us regard the edge $u$ as a morphism of simplicial sets $\Delta ^{1} \rightarrow \operatorname{\mathcal{C}}$, and let $\operatorname{\mathcal{E}}$ denote the fiber product $\Delta ^{1} \times _{\operatorname{\mathcal{C}}} \operatorname{\mathcal{C}}_{/Z}$. Since $q$ is an interior fibration, it follows from Remark 5.4.2.4 and Example 5.4.2.2 that the projection map $\pi : \operatorname{\mathcal{E}}\rightarrow \Delta ^1$ is an inner fibration. Moreover, we can identify $\sigma $ and $\sigma '$ with $\pi $-cartesian edges of $\operatorname{\mathcal{E}}$ having nondegenerate images under $\pi $. Applying Remark 5.1.3.8, we see that there exists a $2$-simplex of $\operatorname{\mathcal{E}}$ which exhibits $\sigma '$ as a composition of $\sigma $ with an isomorphism in $\operatorname{\mathcal{E}}$. The image of this $2$-simplex under the projection map $\operatorname{\mathcal{E}}\rightarrow \operatorname{\mathcal{C}}_{/Z}$ can be identified with a $3$-simplex $\rho $ of $\operatorname{\mathcal{C}}$ such that $d^{3}_0(\rho ) = \sigma $, $d^{3}_1(\rho ) = \sigma '$, and $d^{3}_3(\rho ) = s^{1}_0(u)$ is left-degenerate; the restriction of $\rho $ to the $1$-skeleton of $\Delta ^3$ we can represent by the diagram

\[ \xymatrix@R =50pt@C=50pt{ & X \ar [r]^-{u} \ar [drr]^{w} & Y \ar [dr]^{v} & \\ X \ar [ur]^{\operatorname{id}_ X} \ar [urr]^{ u} \ar [rrr]^{w'} & & & Z. } \]

By construction, the remaining face $\sigma '' = d^{3}_2(\rho )$ is an isomorphism when viewed as a morphism in the $\infty $-category $\operatorname{Hom}^{\mathrm{R}}_{\operatorname{\mathcal{C}}}(X,Z) = \{ X\} \times _{\operatorname{\mathcal{C}}} \operatorname{\mathcal{C}}_{/Z}$, and is therefore locally $q$-cartesian (Example 5.1.3.6). In particular, our inductive hypothesis guarantees that the simplex $\sigma ''$ satisfies condition $(1_{m})$ for $3 \leq m < n$.

Fix a morphism of simplicial sets $\mu _0: \Lambda ^{n}_{i} \rightarrow \operatorname{\mathcal{C}}$ as in condition $(1_ n)$; we wish to show that $\mu _0$ can be extended to an $n$-simplex $\mu $ of $\operatorname{\mathcal{C}}$. Let $\delta ^{i-1}_{n+1}: \Delta ^{n} \hookrightarrow \Delta ^{n+1}$ denote the inclusion of the $(i-1)$st face, given on vertices by the formula

\[ \delta ^{i-1}_{n+1}(j) = \begin{cases} j & \textnormal{ if } j < i-1 \\ j+1 & \textnormal{ if } j \geq i-1. \end{cases} \]

We will construct an $(n+1)$-simplex $\nu : \Delta ^{n+1} \rightarrow \operatorname{\mathcal{C}}$ which satisfies the following conditions:

  • The composite map

    \[ \Lambda ^{n}_{i} \hookrightarrow \Delta ^{n} \xrightarrow { \delta ^{i-1}_{n+1} } \Delta ^{n+1} \xrightarrow {\nu } \operatorname{\mathcal{C}} \]

    is equal to $\mu _0$.

  • The composite map

    \[ \Delta ^3 \simeq \operatorname{N}_{\bullet }( \{ i-1 < i < i+1 < i+2 \} ) \hookrightarrow \Delta ^{n+1} \xrightarrow {\nu } \operatorname{\mathcal{C}} \]

    is equal to the $3$-simplex $\rho $.

  • For every integer $0 \leq j < i-1$, the $2$-simplex

    \[ \Delta ^{2} \simeq \operatorname{N}_{\bullet }( \{ j < i-1 < i \} ) \hookrightarrow \Delta ^{n+1} \xrightarrow {\nu } \operatorname{\mathcal{C}} \]

    is right-degenerate (in particular, it is thin).

  • For every integer $i+2 < j \leq n+1$, the $2$-simplex

    \[ \Delta ^{2} \simeq \operatorname{N}_{\bullet }( \{ i-1 < i < j \} ) \hookrightarrow \Delta ^{n+1} \xrightarrow {\nu } \operatorname{\mathcal{C}} \]

    is left-degenerate (in particular, it is thin).

Assuming that this construction is possible, we complete the proof by observing that $\mu = \nu \circ \delta ^{i-1}_{n+1}$ provides the desired extension of $\mu _0$ (by virtue of assumption $(a)$).

The construction of the $(n+1)$-simplex $\nu $ will take place in several steps. We define simplicial subsets

\[ K_0 \subsetneq K_1 \subsetneq K_2 \subsetneq K_3 \subsetneq K_{4} \subsetneq \Delta ^{n+1} \]

and maps $\nu _ j: K_{j} \rightarrow \operatorname{\mathcal{C}}$ as follows:

  • Let $K_0 \subseteq \Delta ^{n+1}$ be the image of the horn $\Lambda ^ n_{i}$ under $\delta ^{i-1}_{n+1}$, so that $\delta ^{i-1}_{n+1}$ induces an isomorphism $\Lambda ^{n}_{i} \xrightarrow {\sim } K_0$. It follows that there is a unique morphism of simplicial sets $\nu _0: K_0 \rightarrow \operatorname{\mathcal{C}}$ satisfying $\mu _0 = \nu _{0} \circ \delta ^{i-1}_{n+1}|_{ \Lambda ^{n}_{i} }$. By construction, the map $\nu _0$ satisfies condition $(a)$.

  • Let $K_1 \subseteq \Delta ^{n+1}$ be the union of $K_0$ with the $3$-simplex $\operatorname{N}_{\bullet }( \{ i-1 < i < i+1 < i+2 \} )$. It follows from the identity $d^{3}_0(\rho )= \sigma $ that $\nu _0$ extends uniquely to a map $\nu _1: K_1 \rightarrow \operatorname{\mathcal{C}}$ satisfying condition $(b)$.

  • Let $K_2$ be the simplicial subset of $\Delta ^{n+1}$ obtained by removing those nondegenerate simplices which contain all of the vertices $\{ 0 < 1 < \cdots < i-2 < i+2 < i+3 < \cdots < n+1 \} $ and at least one of the vertices $\{ i-1, i \} $. We will prove below that $\nu _1$ can be extended to a map $\nu _2: K_2 \rightarrow \operatorname{\mathcal{C}}$ which satisfies conditions $(c)$ and $(d)$.

  • Let $\delta ^{i}_{n+1}: \Delta ^{n} \hookrightarrow \Delta ^{n+1}$ denote the inclusion of the $i$th face, given on vertices by the formula

    \[ \delta ^{i}_{n+1}(j) = \begin{cases} j & \textnormal{ if } j < i \\ j+1 & \textnormal{ if } j \geq i. \end{cases} \]

    Let $K_3$ be the union of $K_2$ with the image of $\delta ^{i}_{n+1}$. Note that $\delta ^{i}_{n+1}$ determines a pushout diagram of simplicial sets

    \[ \xymatrix@R =50pt@C=50pt{ \Lambda ^{n}_{i} \ar [r] \ar [d] & K_2 \ar [d] \\ \Delta ^{n} \ar [r]^-{\delta ^{i}_{n+1}} & K_3. } \]

    Let $\alpha _0$ denote the composite map $\Lambda ^{n}_{i} \xrightarrow { \delta ^{i}_{n+1} } K_2 \xrightarrow { \nu _2 } \operatorname{\mathcal{C}}$. Since $\nu _1$ satisfies condition $(b)$, $\alpha _0$ carries $\operatorname{N}_{\bullet }( \{ i-1 < i < i+1 \} )$ to the thin $2$-simplex $\sigma '$ of $\operatorname{\mathcal{C}}$, and can therefore be extended to an $n$-simplex $\alpha $ of $\operatorname{\mathcal{C}}$. It follows that $\nu _2$ extends uniquely to a morphism of simplicial sets $\nu _3: K_3 \rightarrow \operatorname{\mathcal{C}}$ satisfying $\nu _3 \circ \delta ^{i}_{n+1} = \alpha $.

  • Let $K_{4}$ denote the horn $\Lambda ^{n+1}_{i-1} \subsetneq \Delta ^{n}$. Note that $K_4$ can be written as the union of $K_3$ with the image of the face inclusion $\delta ^{i+1}_{n+1}: \Delta ^{n} \hookrightarrow \Delta ^{n+1}$, given on vertices by the formula

    \[ \delta ^{i+1}_{n+1}(j) = \begin{cases} j & \textnormal{ if } j \leq i \\ j+1 & \textnormal{ if } j > i. \end{cases} \]

    Moreover, we have a pushout diagram

    \[ \xymatrix@R =50pt@C=50pt{ \Lambda ^{n}_{i-1} \ar [r] \ar [d] & K_3 \ar [d] \\ \Delta ^{n} \ar [r]^-{\delta ^{i+1}_{n+1}} & K_4. } \]

    Let $\beta _0$ denote the composite map

    \[ \Lambda ^{n}_{i-1} \xrightarrow { \delta ^{i+1}_{n+1} } K_3 \xrightarrow { \nu _3 } \operatorname{\mathcal{C}}. \]

    If $i > 1$, then condition $(c)$ guarantees that the restriction $\beta _0|_{\operatorname{N}_{\bullet }( \{ i-2 < i-1 < i \} )}$ is a right-degenerate $2$-simplex of $\operatorname{\mathcal{C}}$. If $i=1$, then condition $(d)$ guarantees that the restriction $\beta _0|_{\operatorname{N}_{\bullet }( \{ 0 < 1 < n \} )}$ is a left-degenerate $2$-simplex of $\operatorname{\mathcal{C}}$. In either case, our assumption that $\operatorname{\mathcal{C}}$ is an $(\infty ,2)$-category guarantees that $\beta _0$ can be extended to an $n$-simplex $\beta $ of $\operatorname{\mathcal{C}}$, so that $\nu _3$ can be extended uniquely to a map $\nu _4: K_4 \rightarrow \operatorname{\mathcal{C}}$ satisfying $\nu _{4} \circ \delta ^{i+1}_{n+1} = \beta $.

  • If $i > 1$, then condition $(c)$ guarantees that the map $\nu _{4}: \Lambda ^{n+1}_{i-1} \rightarrow \operatorname{\mathcal{C}}$ carries $\operatorname{N}_{\bullet }( \{ i-2 < i-1 < i \} )$ to a right-degenerate $2$-simplex of $\operatorname{\mathcal{C}}$. If $i=1$, then condition $(d)$ guarantees that $\nu _{4}$ carries $\operatorname{N}_{\bullet }( \{ 0 < 1 < n+1 \} )$ to a left-degenerate $2$-simplex of $\operatorname{\mathcal{C}}$. In either case, our assumption that $\operatorname{\mathcal{C}}$ is an $(\infty ,2)$-category guarantees that we can extend $\nu _{4}$ to an $(n+1)$-simplex $\nu : \Delta ^{n+1} \rightarrow \operatorname{\mathcal{C}}$, thereby completing the proof of Theorem 5.4.4.1.

It remains to show that $\nu _1$ admits an extension $\nu _2: K_2 \rightarrow \operatorname{\mathcal{C}}$ which satisfies conditions $(c)$ and $(d)$. Let us say that a simplex $\tau : \Delta ^{m} \rightarrow K_2$ is free if it is nondegenerate, not contained in $K_1$, and there exists an integer $0 \leq j \leq m$ satisfying $\tau (j) = i$. Note that in this case, we automatically have $j > 0$ and $\tau (j-1) = i-1$ (otherwise, $\tau $ would be contained in $K_1$). Moreover, if $\tau $ is any nondegenerate $m$-simplex of $K_2$ which is not contained in $K_1$, then $\tau $ is either free or can be realized uniquely as a face of a free $(m+1)$-simplex $\tau ': \Delta ^{m+1} \rightarrow K_2$ (obtained by adjoining $i$ to the image of $\tau $).

Let $\{ \tau _1, \tau _2, \cdots , \tau _ t \} $ be an enumeration of the collection of all free simplices of $K_{2}$, chosen so $\dim ( \tau _1 ) \leq \dim ( \tau _2) \leq \cdots \leq \dim ( \tau _ t )$. For $0 \leq s \leq t$, let $K_{2}(s)$ denote the union of $K_{1}$ with the images of the maps $\{ \tau _1, \tau _2, \cdots , \tau _{s} \} $, so that we have inclusions of simplicial sets

\[ K_1 = K_2(0) \subset K_2(1) \subset K_2(2) \subset \cdots \subset K_2(t) = K_2. \]

We will complete the proof by inductively constructing a compatible sequence of maps $\nu _{2}(s): K_2(s) \rightarrow \operatorname{\mathcal{C}}$ satisfying $\nu _{2}(0) = \nu _1$ together with the following translation of conditions $(c)$ and $(d)$:

$(\ast _{s})$

If the simplex $\tau _{s}$ has dimension $2$, then the $2$-simplex $\nu _{2} \circ \tau _{s}$ of $\operatorname{\mathcal{C}}$ is left-degenerate if $\tau _ s(1) =i$ and right-degenerate if $\tau _{s}(2) = i$.

Assume that $s > 0$ and that the map $\nu _{2}(s-1)$ has already been constructed. Set $\tau = \tau _{s}: \Delta ^{m} \rightarrow K_2$, so that there is a unique integer $1 \leq j \leq m$ satisfying $\tau (j) = i$. Note that for $0 \leq k \leq m$ with $k \neq j$, the face $d^{m}_ k(\tau )$ is either free or belongs to $K_1$; in either case, it belongs to $K_2(s-1)$. Moreover, the face $d^{m}_ j(\tau )$ is neither free, nor contained in $K_1$, nor contained as a face of any other free $m$-simplex of $K_2$. It follows that $\tau $ determines a pushout diagram of simplicial sets

\[ \xymatrix@R =50pt@C=50pt{ \Lambda ^{m}_{j} \ar [r] \ar [d] & K_2(s-1) \ar [d] \\ \Delta ^{m} \ar [r]^-{\tau } & K_2(s). } \]

Let $\xi _0 : \Lambda ^{m}_{j} \rightarrow \operatorname{\mathcal{C}}$ denote the composite map $\Lambda ^{m}_{j} \xrightarrow {\tau } K_{2}(s-1) \xrightarrow { \nu _{2}(s-1)} \operatorname{\mathcal{C}}$; we wish to show that $\xi _0$ can be extended to an $m$-simplex of $\operatorname{\mathcal{C}}$. If $m = 2$, then there is a unique such extension which satisfies condition $(\ast _ s)$ (since, by construction, the morphism $\nu _1$ carries $\operatorname{N}_{\bullet }( \{ i-1 < i \} )$ to the degenerate edge $\operatorname{id}_{X}$ of $\operatorname{\mathcal{C}}$). We may therefore assume that $m \geq 3$. We consider several cases:

  • If $j=m$, then it follows from assumption $(\ast _{s'} )$ for $s' < s$ that $\xi _{0}$ carries $\operatorname{N}_{\bullet }( \{ 0 < m-1 < m \} )$ to a right-degenerate $2$-simplex of $\operatorname{\mathcal{C}}$, so the desired extension exists by virtue of our assumption that $\operatorname{\mathcal{C}}$ is an $(\infty ,2)$-category.

  • If $j < m$ and $\tau (j+1) = i+1$, then it follows from $(b)$ that $\xi _0$ carries $\operatorname{N}_{\bullet }( \{ j-1 < j < j+1 \} )$ to the left-degenerate $2$-simplex $d^{3}_3(\rho )$. Since $\operatorname{\mathcal{C}}$ is an $(\infty ,2)$-category, this $2$-simplex is thin so that $\xi _0$ can be extended to an $m$-simplex of $\operatorname{\mathcal{C}}$.

  • If $j < m$ and $\tau (j+1) > i+2$, then it follows from assumption $(\ast _{s'})$ for $s' < s$ that $\xi _0$ carries $\operatorname{N}_{\bullet }( \{ j-1 < j < j+1 \} )$ to a left-degenerate $2$-simplex of $\operatorname{\mathcal{C}}$. Since $\operatorname{\mathcal{C}}$ is an $(\infty ,2)$-category, this $2$-simplex is thin so that $\xi _0$ can be extended to an $m$-simplex of $\operatorname{\mathcal{C}}$.

  • If $j < m$ and $\tau (j+1) = i+2$, then it follows from $(b)$ that $\xi _{0}$ carries $\operatorname{N}_{\bullet }( \{ j-1 < j < j+1 \} )$ to the $2$-simplex $\sigma ''$ of $\operatorname{\mathcal{C}}$. In this case, our assumption that $\tau $ belongs to $K_2$ guarantees that $m < n$, so the existence of the desired extension follows the fact that $\sigma ''$ satisfies condition $(1_{m})$ (by virtue of our inductive hypothesis).

$\square$