1:
2: \defnsheading{L} \label{p:l}
3:
4:
5: \concept{Globular Operads and their Algebras}
6:
7: \paragraph{Globular Sets}
8:
9: Let \scat{G} be the category whose objects are the natural numbers
10: $0,1,\ldots$, and whose arrows are generated by
11: $
12: \sigma_m, \tau_m: m \go m-1
13: $
14: for each $m\geq 1$, subject to equations
15: \[
16: \sigma_{m-1} \of \sigma_m = \sigma_{m-1} \of \tau_m,
17: \diagspace
18: \tau_{m-1} \of \sigma_m = \tau_{m-1} \of \tau_m
19: \]
20: ($m\geq 2$). A functor $A: \scat{G} \go \Set$ is called a \demph{globular
21: set}; I will write $s$ for $A(\sigma_m)$, and $t$ for $A(\tau_m)$.
22:
23: \paragraph{The Free Strict $\omega$-Category Monad}
24:
25: Any
26: (small)
27: strict $\omega$-category has an underlying globular set $A$, in
28: which $A(m)$ is the set of $m$-cells and $s$ and $t$ are the source and
29: target maps. We thus obtain a forgetful functor $U$ from the category of
30: strict $\omega$-categories and strict $\omega$-functors to the category
31: \ftrcat{\scat{G}}{\Set} of globular sets. $U$ has a left adjoint, so there
32: is an induced monad $(T, \id \goby{\eta} T, T^2 \goby{\mu} T)$ on
33: \ftrcat{\scat{G}}{\Set}.
34:
35: \paragraph{Collections}
36:
37: We define a monoidal category \fcat{Coll} of collections. Let $1$ be the
38: terminal globular set. A \demph{(globular) collection} is a map $C \goby{d}
39: T1$ into $T1$ in \ftrcat{\scat{G}}{\Set}; a \demph{map of collections} is a
40: commutative triangle. The \demph{tensor product} of collections $C \goby{d}
41: T1$, $C' \goby{d'} T1$ is the composite along the top row of
42: %
43: \begin{diagram}[size=2em]
44: \SEpbk C \otimes C'&\rTo&TC' &\rTo^{Td'} &T^2 1 &\rTo^{\mu_1}&T1\\
45: \dTo & &\dTo>{T!}& & & & \\
46: C &\rTo^d &T1, & & & & \\
47: \end{diagram}
48: %
49: where the right-angle symbol means that the square containing it is a
50: pullback, and $!$ denotes the unique map to $1$. The \demph{unit} for the
51: tensor is $1 \goby{\eta_1} T1$.
52:
53: \paragraph{Globular Operads}
54:
55: A \demph{(globular) operad} is a monoid in the monoidal category \fcat{Coll};
56: a \demph{map of operads} is a map of monoids.
57:
58: \paragraph{Algebras}
59:
60: Any operad $C$ induces a monad $C\cdot\dashbk$ on \ftrcat{\scat{G}}{\Set}.
61: For an object $A$ of \ftrcat{\scat{G}}{\Set}, this is defined by pullback:
62: %
63: \begin{diagram}[size=2em]
64: \SEpbk C\cdot A &\rTo &TA \\
65: \dTo & &\dTo>{T!} \\
66: C &\rTo^d &T1. \\
67: \end{diagram}
68: %
69: The multiplication and unit of the monad come from the multiplication and
70: unit of the operad. A \emph{$C$-algebra} is an algebra for the monad
71: $C\cdot\dashbk$. Note that every $C$-algebra has an underlying globular
72: set.
73:
74:
75: \concept{Contractions}
76:
77: \paragraph{Contractions}
78:
79: Let $C \goby{d} T1$ be a collection. For $m\geq 0$ and $\nu\in (T1)(m)$,
80: write
81: $
82: C(\nu) = \{ \theta \in C(m) \such d(\theta) = \nu \}.
83: $
84: For $m\geq 2$ and $\pi\in (T1)(m)$, define
85: \[
86: P_C(\pi) =
87: \{ \pr{\theta_0}{\theta_1} \in C(s(\pi)) \times C(t(\pi)) \such
88: s(\theta_0) = s(\theta_1) \mbox{ and } t(\theta_0) = t(\theta_1)\},
89: \]
90: and for $\pi\in (T1)(1)$, define
91: $
92: P_C(\pi) = C(s(\pi)) \times C(t(\pi)).
93: $
94: A \demph{contraction} $\gamma$ on $C$ is a family of functions
95: \[
96: (\gamma_\pi: P_C(\pi) \go C(\pi))_{m\geq 1, \pi\in (T1)(m)}
97: \]
98: satisfying
99: \[
100: s(\gamma_\pi \pr{\theta_0}{\theta_1} ) = \theta_0,
101: \diagspace
102: t(\gamma_\pi \pr{\theta_0}{\theta_1} ) = \theta_1
103: \]
104: for every $m\geq 1$, $\pi\in (T1)(m)$ and $\pr{\theta_0}{\theta_1} \in
105: P_C(\pi)$.
106:
107: \paragraph{Initial Object}
108:
109: Let \fcat{OC} be the category in which an object is an operad equipped with a
110: contraction and a map is a map of operads preserving the specified
111: contraction. Then \fcat{OC} can be shown to have an initial object, whose
112: underlying operad will be written $L$.
113:
114:
115:
116: \concept{The Definitions}
117:
118: \paragraph{Definition \ds{L1}}
119:
120: A \demph{weak $\omega$-category} is an $L$-algebra. (Maps of
121: $L$-algebras should be regarded as \emph{strict} $\omega$-functors.)
122:
123: \paragraph{Definition \ds{L2}}
124:
125: A \demph{weak $\omega$-category} is a pair $(C,A)$, where $C$ is an operad on
126: which there exists a contraction and satisfying $C(0) \iso 1$, and $A$ is a
127: $C$-algebra.
128:
129: \paragraph{Weak $n$-Categories}
130:
131: Let $n\geq 0$. A globular set $A$ is \emph{$n$-dimensional} if for all
132: $m\geq n$,
133: \[
134: s=t: A(m+1) \go A(m)
135: \]
136: and this map is an isomorphism. A \demph{weak $n$-category} is a weak
137: $\omega$-category whose underlying globular set is $n$-dimensional. This can
138: be interpreted according to either \ds{L1} or \ds{L2}.
139:
140:
141: \clearpage
142:
143:
144:
145: \lowsdimsheading{L}
146:
147: \concept{Definition \ds{L1}}
148:
149: An alternative way of handling weak $n$-categories is to work with only $n$-
150: (not infinite-) dimensional structures throughout. So we replace $\scat{G}$
151: by its full subcategory $\scat{G}_n$ with objects $0, \ldots, n$, and $T$ by
152: the free strict $n$-category monad $T_n$, to obtain definitions of
153: \demph{$n$-collection}, \demph{$n$-operads}, and their \demph{algebras}.
154: \demph{Contractions} are defined as before, except that we only speak of
155: contractions on $C$ if
156: %
157: \begin{equation} \label{eq:n-contr-l}
158: \forall \nu \in (T_n 1)(n), \forall \theta_0, \theta_1 \in C(\nu), \
159: s(\theta_0) = s(\theta_1) \ \&\ t(\theta_0) = t(\theta_1) \implies
160: \theta_0 = \theta_1
161: \end{equation}
162: %
163: (taking $C(-1)=1$ to understand this when $n=0$). There is an initial
164: $n$-operad $L_n$ equipped with a contraction, and the category of weak
165: $n$-categories turns out to be equivalent to the category of $L_n$-algebras.
166: The latter is easier to analyse.
167:
168: \paragraph{$n=0$}
169:
170: We have $\ftrcat{\scat{G}_0}{\Set} \iso \Set$, $T_0 = \id$, and
171: $0\hyph\fcat{Coll} \iso \Set$; a 0-operad $C$ is a monoid, and a $C$-algebra
172: is a set with a $C$-action. By~\bref{eq:n-contr-l}, the only
173: 0-operad with a contraction is the one-element monoid, so a weak $0$-category
174: is just a set.
175:
176: \paragraph{$n=1$}
177:
178: $\ftrcat{\scat{G}_1}{\Set}$ is the category of directed graphs and $T_1$ is
179: the free category monad. $L_1$ is the terminal 1-operad, by arguments
180: similar to those under `$n=2$' below. It follows that the induced monad $L_1
181: \cdot \dashbk$ is just $T_1$, and so a weak $1$-category is just a
182: $T_1$-algebra, that is, a category.
183:
184: \paragraph{$n=2$}
185:
186: A functor $A: \scat{G}_2 \go \Set$ consists of a set of $0$-cells (drawn
187: $\gzeros{a}$), a set of $1$-cells ($\gfsts{a}\gones{f}\glsts{b}$), and a set
188: of $2$-cells ($\gfsts{a}\gtwos{f}{g}{\alpha}\glsts{b}$). A 2-collection $C$
189: consists of a set $C(0)$, a set $C(\nu_k)$ for each $k\geq 0$ (where $\nu_k$
190: indicates the `1-pasting diagram' $\gfstsu\gonesu\ \ldots\ \gonesu\glstsu$
191: with $k$ arrows), and a set $C(\pi)$ for each `2-pasting diagram' $\pi$ such
192: as the $\pi_i$ in Fig.~\ref{fig:op-comp-l}, together with source and target
193: functions.
194:
195: A 2-operad is a 2-collection $C$ together with `composition' functions such as
196: \[
197: \begin{array}{ccc}
198: C(\nu_3) \times [C(\nu_2) \times_{C(0)} C(\nu_1) \times_{C(0)} C(\nu_2)]
199: &\go &
200: C(\nu_5), \\
201: C(\pi_1) \times [C(\pi_2) \times_{C(\nu_2)} C(\pi_3)] &\go & C(\pi_4).
202: \end{array}
203: \]
204: In the first, the point is that there are $3$ terms $2,1,2$ and their sum is
205: $5$. This makes sense if an element of $C(\nu_k)$ is regarded as an
206: operation which takes a string of $k$ 1-cells and turns it into a single
207: $1$-cell. (The $\times_{C(0)}$'s denote pullbacks.) Similarly for the
208: second; see Fig.~\ref{fig:op-comp-l}. There are also identities for the
209: compositions.
210: %
211: \begin{figure}
212: \piccy{compoppic.ps}
213: \caption{Composition of operations in a globular operad}
214: \label{fig:op-comp-l}
215: \end{figure}
216: %
217: A $C$-algebra is a functor $A: \scat{G}_2 \go \Set$ together with functions
218: \[
219: \begin{array}{l}
220: \ovln{\psi}: A(0) \go A(0)
221: \textrm{ for each } \psi \in C(0),
222: \\
223: \ovln{\phi}: \{ \textrm{diagrams } \gfsts{a_0}\gones{f_1}\ \cdots \
224: \gones{f_k} \glsts{a_k} \textrm{ in } A \} \go A(1)
225: \textrm{ for each } \phi \in C(\nu_k),
226: \\
227: \ovln{\theta}: \{ \textrm{diagrams }
228: \gfsts{a}%
229: \gthrees{f}{g}{h}{\alpha}{\beta}%
230: \gfbws{\ b}\gtwos{l}{m}{\!\gamma}\glsts{c}
231: \textrm{ in } A \} \go A(2)
232: \textrm{ for each } \theta \in C(\gfstsu\gthreesu\gzersu\gtwosu\glstsu)
233: \end{array}
234: \]
235: (etc), all compatible with the source, target, composition and identities in
236: $C$.
237:
238: $L_2$ is generated from the empty collection by adding in the minimal amount
239: to obtain a 2-operad-with-contraction. We have the identity $1\in L_2(0)$.
240: Then, contraction gives an element of $L_2(\nu_k)$ for each $k$, so that
241: composition in $L_2$ gives an element of $L_2(\nu_k)$ for each $k$-leafed
242: tree. Contraction at the next level (with~\bref{eq:n-contr-l}) says that if
243: $\pi$ is a 2-pasting diagram of width $k$ then $L_2(\pi) = L_2(\nu_k) \times
244: L_2(\nu_k)$. So $L_2(0) = \{1\}$, $L_2(\nu_k) = \{ k \textrm{-leafed trees}
245: \}$, $L_2(\pi) = \{ k \textrm{-leafed trees} \}^2$. An $L_2$-algebra is,
246: then, an `unbiased bicategory': that is, just like a bicategory except that
247: there is specified $k$-fold composition for every $k\geq 0$ rather than just
248: $k=2$ (binary composition) and $k=0$ (identities). Since these are
249: essentially the same as ordinary bicategories, so too are weak
250: $2$-categories.
251:
252:
253: \concept{Definition \ds{L2}}
254:
255: Definition \ds{L2} of weak $n$-category refers to infinite-dimensional
256: globular operads. So in order to do a concrete analysis of $n\leq 2$, we
257: redefine a \demph{weak $n$-category} as a pair $(C,A)$ where $C$ is a
258: $n$-operad admitting a contraction and with $C(0) \iso 1$, and $A$ is a
259: $C$-algebra. (Temporarily, call such a $C$ \demph{good}.) I do not know to
260: what extent this is equivalent to \ds{L2}, but the spirit, at least, is the
261: same.
262:
263:
264: \paragraph{$n=0$}
265:
266: From `$n=0$' above we see that a weak $0$-category is just a set.
267:
268:
269: \paragraph{$n=1$}
270:
271: The only good $1$-operad is the terminal $1$-operad, so by `$n=1$' above, a
272: weak $1$-category is just a category.
273:
274:
275: \paragraph{$n=2$}
276:
277: A \demph{(non-symmetric) classical operad} $D$ is a sequence $(D(k))_{k\geq
278: 0}$ of sets together with an element (the \demph{identity}) of $D(1)$ and for
279: each $k, r_1, \ldots, r_k \geq 0$ a map $ D(k) \times D(r_1) \times \cdots
280: \times D(r_k) \go D(r_1 + \cdots + r_k) $ (\demph{composition}), obeying unit
281: and associativity laws. It turns out that good $2$-operads $C$ correspond
282: one-to-one with classical operads $D$ such that $D(k) \neq \emptyset$ for
283: each $k$, via $D(k) = C(\nu_k)$. A $C$-algebra is then something like a
284: 2-category or bicategory, with one way of composing a string of $k$ $1$-cells
285: for each element of $D(k)$, and all the appropriate coherence $2$-cells.
286: E.g.\ if $D=1$ then a $C$-algebra is a 2-category; if $D(k)$ is the set of
287: $k$-leafed trees in which each vertex has either $0$ or $2$ edges coming up
288: out of it then a $C$-algebra is a bicategory. $C$ can, therefore, be
289: regarded as a theory of (more or less weak) 2-categories, and $A$ as a model
290: for such a theory.
291:
292:
293:
294:
295:
296:
297:
298:
299:
300:
301:
302:
303:
304:
305: