2.3.5 The Duskin Nerve of a Strict $2$-Category
Let $\operatorname{\mathcal{C}}$ be a strict $2$-category (Definition 2.2.0.1). Then we can regard $\operatorname{\mathcal{C}}$ as a $2$-category (in which the associativity and unit constraints are identity morphisms), and form the Duskin nerve $\operatorname{N}_{\bullet }^{\operatorname{D}}(\operatorname{\mathcal{C}})$ by applying Construction 2.3.1.1. However, the Duskin nerve of a strict $2$-category admits a more direct description, which can be formulated entirely in terms of strict $2$-categories (and strict functors between them). The proof is based on a construction which will play an important role in §2.4.3.
Construction 2.3.5.1 (The Path $2$-Category of a Partially Ordered Set). Let $(Q, \leq )$ be a partially ordered set. We define a strict $2$-category $\operatorname{Path}_{(2)}[Q]$ as follows:
The objects of $\operatorname{Path}_{(2)}[Q]$ are the elements of $Q$.
Given elements $x,y \in Q$, we let $\underline{\operatorname{Hom}}_{ \operatorname{Path}[Q]_{(2)} }(x, y)$ denote the partially ordered set of all finite linearly ordered subsets
\[ S = \{ x = x_0 < x_1 < \cdots < x_ n = y \} \subseteq Q \]
having least element $x$ and greatest element $y$, ordered by reverse inclusion. We regard the partially ordered set $\underline{\operatorname{Hom}}_{\operatorname{Path}[Q]_{(2)}}(x,y)$ as a category, having a unique morphism $S \Rightarrow T$ when $T$ is contained in $S$.
For every element $x \in Q$, the identity $1$-morphism $\operatorname{id}_{x} \in \underline{\operatorname{Hom}}_{\operatorname{Path}[Q]}( x, x)$ is given by the singleton $\{ x\} $ (regarded as a linearly ordered subset of $Q$, having greatest and least element $x$).
For every triple of objects $x,y,z \in Q$, the composition functor
\[ \circ : \underline{\operatorname{Hom}}_{\operatorname{Path}_{(2)}[Q]}(y,z) \times \underline{\operatorname{Hom}}_{\operatorname{Path}_{(2)}[Q]}(x,y) \rightarrow \underline{\operatorname{Hom}}_{\operatorname{Path}_{(2)}[Q]}(x,z) \]
is given on objects by the construction $(S, T) \mapsto S \cup T$.
‘ We will refer to $\operatorname{Path}_{(2)}[Q]$ as the path $2$-category of $Q$.
Note that $\operatorname{Path}[Q]$ can also be realized as the path category of a directed graph $\mathrm{Gr}(Q)$ (as defined in Construction 1.3.7.1). Here $\mathrm{Gr}(Q)$ denotes the underlying directed graph of the category $Q$, given concretely by
\[ \operatorname{Vert}( \mathrm{Gr}(Q) ) = Q \quad \quad \operatorname{Edge}( \mathrm{Gr}(Q) ) = \{ (x,y) \in Q: x < y \} \]
where we regard each ordered pair $(x,y) \in \operatorname{Edge}( \mathrm{Gr}(Q) )$ as an edge with source $s(x,y) = x$ and target $t(x,y) = y$.
Construction 2.3.5.4. Let $(Q, \leq )$ be a partially ordered set, which we regard as a category having a unique morphism $e_{y,x}$ for every pair of elements $x,y \in Q$ with $x \leq y$. We define a strictly unitary lax functor $T_{Q}: Q \rightarrow \operatorname{Path}_{(2)}[Q]$ as follows:
On objects, the lax functor $T_{Q}$ is given by $T_{Q}(x) = x$.
On $1$-morphisms, the lax functor $T_{Q}$ is given by $T_{Q}( e_{y,x} ) = \{ y,x \} \in \underline{\operatorname{Hom}}_{ \operatorname{Path}_{(2)}[Q]}(x,y)$ whenever $x \leq y$ in $Q$.
For every triple of elements $x,y,z \in Q$ satisfying $x \leq y \leq z$, the composition constraint $\mu _{z,y,x}: T_{Q}( e_{z,y} ) \circ T_{Q}( e_{y,z} ) \Rightarrow T_{Q}( e_{z,x} )$ is the $2$-morphism of $\operatorname{Path}_{(2)}[Q]$ corresponding to the inclusion of linearly ordered sets
\[ T_ Q( e_{z,x} ) = \{ z, x \} \subseteq \{ z, y, x \} = \{ z, y \} \cup \{ y, x \} = T_{Q}( e_{z,y} ) \circ T_{Q}( e_{y,x} ). \]
The $2$-category $\operatorname{Path}_{(2)}[Q]$ of Construction 2.3.5.1 is characterized by the following universal property:
Theorem 2.3.5.6. Let $Q$ be a partially ordered set and let $T_{Q}: Q \rightarrow \operatorname{Path}_{(2)}[Q] $ be the lax functor of Construction 2.3.5.4. For every strict $2$-category $\operatorname{\mathcal{C}}$, composition with $T_{Q}$ induces a bijection
\[ \{ \textnormal{Strict functors $F^{+}: \operatorname{Path}_{(2)}[Q] \rightarrow \operatorname{\mathcal{C}}$} \} \rightarrow \{ \textnormal{Strictly unitary lax functors $F: Q \rightarrow \operatorname{\mathcal{C}}$} \} . \]
Before giving the proof of Theorem 2.3.5.6, let us note one of its consequences. The construction $[n] \mapsto \operatorname{Path}_{(2)}[n]$ determines a functor from the simplex category $\operatorname{{\bf \Delta }}$ of Definition 1.1.0.2 to the (ordinary) category $\operatorname{2Cat}_{\operatorname{Str}}$ of strict $2$-categories (Definition 2.2.5.5). We will view this functor as a cosimplicial object of $\operatorname{2Cat}_{\operatorname{Str}}$ which we denote by $\operatorname{Path}_{(2)}[\bullet ]$. Applying the construction of Variant 1.2.2.8, we obtain a functor $\operatorname{Sing}_{\bullet }^{ \operatorname{Path}_{(2)}[\bullet ] }: \operatorname{2Cat}_{\operatorname{Str}} \rightarrow \operatorname{Set_{\Delta }}$, which carries each strict $2$-category $\operatorname{\mathcal{C}}$ to the simplicial set $[n] \mapsto \operatorname{Hom}_{ \operatorname{2Cat}_{\operatorname{Str}} }( \operatorname{Path}_{(2)}[\bullet ], \operatorname{\mathcal{C}})$. Using Theorem 2.3.5.6, we can identify this construction with the Duskin nerve functor
\[ \operatorname{2Cat}_{\operatorname{Str}} \hookrightarrow \operatorname{2Cat}_{\operatorname{ULax}} \xrightarrow { \operatorname{N}_{\bullet }^{\operatorname{D}} } \operatorname{Set_{\Delta }}. \]
In particular, we have the following:
Corollary 2.3.5.7. For every strict $2$-category $\operatorname{\mathcal{C}}$, there is a canonical isomorphism of simplicial sets
\[ \operatorname{Sing}_{\bullet }^{\operatorname{Path}_{(2)}[\bullet ] }(\operatorname{\mathcal{C}}) \simeq \operatorname{N}^{\operatorname{D}}_{\bullet }(\operatorname{\mathcal{C}}), \]
given on $n$-simplices by composition with the lax functor $T_{[n]}: [n] \rightarrow \operatorname{Path}[n]$ of Construction 2.3.5.4. In other words, the Duskin nerve $\operatorname{N}^{\operatorname{D}}_{\bullet }(\operatorname{\mathcal{C}})$ is given by
\[ \operatorname{N}^{\operatorname{D}}_{n}(\operatorname{\mathcal{C}}) \simeq \{ \textnormal{Strict functors $\operatorname{Path}_{(2)}[n] \rightarrow \operatorname{\mathcal{C}}$} \} . \]
Proof of Theorem 2.3.5.6.
Let $\operatorname{\mathcal{C}}$ be a strict $2$-category, let $Q$ be a partially ordered set, and let $F: Q \rightarrow \operatorname{\mathcal{C}}$ be a strictly unitary lax functor. We wish to show that $F$ factors uniquely as a composition
\[ Q \xrightarrow {T_ Q} \operatorname{Path}_{(2)}[Q] \xrightarrow {F^{+} } \operatorname{\mathcal{C}}, \]
where $T_{Q}$ is the strictly unitary lax functor of Construction 2.3.5.4 and $F^{+}$ is a strict functor from $\operatorname{Path}_{(2)}[Q]$ to $\operatorname{\mathcal{C}}$.
For every pair of elements $x,y \in Q$ satisfying $x \leq y$, we let $e_{y,x}: x \rightarrow y$ denote the unique morphism from $x$ to $y$ in the category $Q$, and for every triple $x,y,z \in Q$ satisfying $x \leq y \leq z$, we let $\mu _{z,y,x}: F( e_{z,y} ) \circ F( e_{y,x} ) \Rightarrow F( e_{z,x} )$ denote the composition constraint for the lax monoidal functor $F$. Unwinding the definitions, we see that a strict functor $F^{+}: \operatorname{Path}_{(2)}[Q] \rightarrow \operatorname{\mathcal{C}}$ satisfies $F^{+} \circ T_{Q} = F$ if and only if the following conditions are satisfied:
- $(0)$
For every element $x \in Q$, we have $F^{+}(x) = F(x)$ (as objects of the $2$-category $\operatorname{\mathcal{C}}$).
- $(1)$
For every pair of elements $x,y \in Q$ satisfying $x \leq y$, we have $F^{+}( \{ y, x\} ) = F( e_{y,x} )$ (as $1$-morphisms from $F(x)$ to $F(y)$ in the strict $2$-category $\operatorname{\mathcal{C}}$).
- $(2)$
For every triple of elements $x,y,z \in Q$ satisfying $x \leq y \leq z$, the functor $F^{+}$ carries the inclusion $\{ z, x \} \subseteq \{ z,y,x \} $ (regarded as a $2$-morphism from $\{ z, y\} \circ \{ y,x\} $ to $\{ z, x\} $ in the strict $2$-category $\operatorname{Path}_{(2)}[Q]$) to $\mu _{z,y,x}$ (regarded as a $2$-morphism from $F( e_{z,y} ) \circ F( e_{ y,x} )$ to $F( e_{z,x} )$ in the strict $2$-category $\operatorname{\mathcal{C}}$).
Note that, since we are requiring $F^{+}$ to be a strict functor, we can replace $(1)$ by the following stronger condition:
- $(1')$
For every nonempty finite linearly ordered subset $S = \{ x_0 < x_1 < \cdots < x_ n \} \subseteq Q$, the functor $F^{+}$ carries $S$ (regarded as a $1$-morphism from $x_0$ to $x_ n$ in the strict $2$-category $\operatorname{Path}_{(2)}[Q]$) to the composition $F( e_{ x_{n}, x_{n-1} } ) \circ \cdots \circ F( e_{ x_{1}, x_0} )$ (regarded as a $1$-morphism from $F( x_0)$ to $F(x_ n)$ in the strict $2$-category $\operatorname{\mathcal{C}}$). In what follows, we will denote this composition by $F(S)$.
Let $S = \{ x_0 < x_1 < \cdots < x_ n \} $ be a nonempty finite linearly ordered subset of $Q$. For each $0 \leq i \leq j \leq n$, set $f_{j,i} = F( e_{ x_ j, x_ i} )$, which we regard as a $1$-morphism from $F( x_ i )$ to $F(x_ j)$ in the $2$-category $\operatorname{\mathcal{C}}$. Let $x_ i$ be an element of $S$ which is neither the largest nor the smallest (so that $0 < i < n$). In this case, we let $\gamma _{S, x_ i}: F( S ) \Rightarrow F(S \setminus \{ x_ i \} )$ denote the $2$-morphism of $\operatorname{\mathcal{C}}$ given by the horizontal composition
\[ \gamma _{S,x_ i} = \operatorname{id}_{ f_{n,n-1} } \circ \cdots \circ \operatorname{id}_{ f_{i+2,i+1} } \circ \mu _{ x_{i+1}, x_{i}, x_{i-1} } \circ \operatorname{id}_{ f_{i-1, i-2} } \circ \cdots \circ \operatorname{id}_{ f_{1,0} }. \]
More generally, given a sequence of distinct elements $s_1, s_2, \cdots , s_ m \in S \setminus \{ x_0, x_ n \} $, we let $\gamma _{ S, s_1, \ldots , s_ m }: F(S) \Rightarrow F( S \setminus \{ s_1, \ldots , s_ m \} )$ denote the $2$-morphism of $\operatorname{\mathcal{C}}$ given by the vertical composition
\[ F(S) \xRightarrow { \gamma _{S, s_1} } F(S \setminus \{ s_1\} ) \xRightarrow { \gamma _{ S \setminus \{ s_1\} , s_2} } F( S \setminus \{ s_1, s_2\} ) \Rightarrow \cdots \Rightarrow F( S \setminus \{ s_1, \ldots , s_ m \} ). \]
Since the strict functor $F^{+}$ is required to be compatible with vertical and horizontal composition, we can replace $(2)$ by the following stronger condition:
- $(2')$
Let $S = \{ x_0 < x_1 < \cdots < x_ n \} $ be a nonempty finite linearly ordered subset of $Q$. Then, for every sequence of distinct elements $s_1, \ldots , s_ m \in S \setminus \{ x_0, x_ n \} $, the functor $F^{+}$ carries the inclusion $S \setminus \{ s_1, \ldots , s_ m \} \subseteq S$ (regarded as a $2$-morphism from $S$ to $S \setminus \{ s_1, \ldots , s_ m \} $ in the strict $2$-category $\operatorname{Path}_{(2)}[Q]$) to the $2$-morphism $\gamma _{S, s_1, \ldots , s_ m }$ (regarded as a $2$-morphism from $F(S)$ to $F( S \setminus \{ s_1, \ldots , s_ m \} )$ in the strict $2$-category $\operatorname{\mathcal{C}}$).
It is now clear that the functor $F^{+}$ is unique if it exists: its values on objects, $1$-morphisms, and $2$-morphisms of $\operatorname{Path}_{(2)}[Q]$ are determined by conditions $(0)$, $(1')$, and $(2')$, respectively. To prove existence, it will suffice to show that this prescription is well-defined: namely, that the $2$-morphism $\gamma _{S, s_1, \ldots , s_ m}$ defined above depends only on the sets $S$ and $T = S \setminus \{ s_1, \ldots , s_ m \} $, and not on the order of the sequence $(s_1, \ldots , s_ m )$ (it then follows easily from the construction that the definition of $F^{+}$ on $2$-morphisms is compatible with vertical and horizontal composition). Since the group of all permutations of the set $\{ s_1, \ldots , s_ m \} $ is generated by transpositions of adjacent elements, it will suffice to show that we have
\[ \gamma _{S, s_1, \cdots , s_{i-1}, s_{i}, s_{i+1}, s_{i+2}, \cdots , s_ m } = \gamma _{S, s_1, \cdots , s_{i-1}, s_{i+1}, s_{i}, s_{i+2}, \cdots , s_ m } \]
for each $1 \leq i < m$. Replacing $S$ by $S \setminus \{ s_1, \ldots , s_{i-1} \} $, we are reduced to proving that $\gamma _{S, s, t} = \gamma _{ S, t, s }$ whenever $s < t$ are elements of $S - \{ x_0, x_ n \} $. We now distinguish two cases:
Suppose that the elements $s$ and $t$ are non-consecutive elements of $S$: that is, we have $s = x_ i$ and $t = x_ j$ where $j > i + 1$. In this case, we can identify both $\gamma _{S, s,t}$ and $\gamma _{S, t, s}$ with the horizontal composition
\[ \operatorname{id}_{ f_{n,n-1} } \circ \cdots \circ \mu _{ x_{j+1}, x_{j}, x_{j-1} } \circ \cdots \circ \mu _{ x_{i+1}, x_ i, x_{i-1} } \circ \cdots \circ \operatorname{id}_{ f_{1,0} }. \]
Suppose that the elements $s$ and $t$ are consecutive: that is, we have
\[ S = \{ x_0 < \cdots < r < s < t < u < \cdots < x_ n \} . \]
In this case, to verify the identity $\gamma _{S, s, t} = \gamma _{S, t, s}$, we can replace $S$ by the subset $\{ r < s < t < u \} $ and thereby reduce to checking the commutativity of the diagram
\[ \xymatrix@C =80pt@R=50pt{ F( e_{ u, t} ) \circ F( e_{t,s} ) \circ F( e_{s,r} ) \ar@ {=>}[r]^-{ \operatorname{id}_{ F( e_{u,t} )} \circ \mu _{t,s,r} } \ar@ {=>}[d]^-{ \mu _{u,t,s} \circ \operatorname{id}_{ F( e_{s,r})} } & F( e_{u,t} ) \circ F( e_{ t, r} ) \ar@ {=>}[d]^{ \mu _{u,t,r} } \\ F( e_{u,s} ) \circ F( e_{s,r} ) \ar@ {=>}[r]^{ \mu _{ u,s,r} } & F( e_{u,r} ) } \]
in the category $\underline{\operatorname{Hom}}_{\operatorname{\mathcal{C}}}( F(r), F(u) )$, which is the coherence condition required by the composition contraints for the lax functor $F$ (axiom $(c)$ of Definition 2.2.4.5).
$\square$