Monads with arities and their associated theories (Q456883)

In the article under review, as the authors say in the abstract, after introducing the concept of ``monad with arities'', show in Theorem 1.10 (Nerve Theorem), p. 2033, that the category of algebras for such a monad has a canonical dense generator (see below). This is used to extend the well-known correspondence between finitary monads on the category \({\mathcal Set}\) of sets and Lawvere's algebraic theories (as well as the correspondence between finitary monads on powers of \({\mathcal Set}\) and Bénabou's algebraic theories, I add) to a general correspondence between monads and theories for a given category with arities. Moreover, as an application, the authors determine arities for the free groupoid monad on involutive graphs and recover the Grothendieck's symmetric simplicial nerve characterization of groupoids. Next we proceed to explain more fully the contents of the article under consideration. In Section 1, the authors, after recalling the concept of monad morphism of \textit{R. Street} [J. Pure Appl. Algebra 2, 149--168 (1972; Zbl 0241.18003)], the essential image-factorization of a functor, the notion of isofibration, the concept of monadic functor, and that of exact adjoint square, extend a classical proposition about squares induced by monad morphisms to exact adjoint squares. Next they continue with the definition of the notion of dense generator of a category \(\mathcal{E}\), i.e., a pair \((\mathcal{A},i_{\mathcal{A}})\) (or simply \(\mathcal{A}\)), where \(\mathcal{A}\) is a small full subcategory of \(\mathcal{E}\) and \(i_{\mathcal{A}}\) the canonical embedding of \(\mathcal{A}\) into \(\mathcal{E}\), such that the associated nerve functor \(\nu_{\mathcal{A}}\) from \(\mathcal{E}\) to \(\widehat{\mathcal{A}} = {\mathcal Set}^{\mathcal A^{\mathrm {op}}}\) is fully faithful (for every \(X\) in \(\mathcal{E}\), \(\nu_{\mathcal{A}}(X)\) is the contravariant functor \(\text{Hom}_{\mathcal{E}}(i_{\mathcal{A}}(\cdot),X)\)). After this, for a dense generator \((\mathcal{A},i_{\mathcal{A}})\) of \(\mathcal{E}\), they define a monad with arities \(\mathcal{A}\) as a monad \(\mathbb{T} = (T,\mu,\eta)\) on \(\mathcal{E}\) such that \(\nu_{\mathcal{A}}\circ T\) takes the \(\mathcal{A}\)-cocones in \(\mathcal{E}\) to colimit-cocones in \(\widehat{\mathcal{A}}\). Moreover, the authors give useful characterizations of the concepts of dense generator and of monad with arities (this last one in terms of exact adjoint squares). Following this, they state, in Theorem 1.10 (nerve theorem) (which turns out to be the main tool of the article to develop a framework in which to state several naturally arising monad/theory equivalences) the following: ``Let \(\mathcal{E}\) be a category with dense generator \((\mathcal{A},i_{\mathcal{A}})\). For any monad \(\mathbb{T}\) with arities \(\mathcal{A}\), the full subcategory \(\Theta_{\mathbb{T}}\) spanned by the free \(\mathbb{T}\)-algebras on the arities is a dense generator of the Eilenberg-Moore category \(\mathcal{E}^{\mathbb{T}} = {\mathcal EM}(\mathbb{T})\). The essential image of the nerve functor \(\nu_{\mathbb{T}}\) from \({\mathcal EM}(\mathbb{T})\) to \(\widehat{\Theta}_{\mathbb{T}}\) is spanned by those presheaves whose restriction along \(j_{\mathbb{T}}\) belongs to the essential image of \(\nu_{\mathcal{A}}\)'' (where, I recall, \(\Theta_{\mathbb{T}}\) is the full subcategory of \({\mathcal EM}(\mathbb{T})\) spanned by the free \(\mathbb{T}\)-algebras on the objects of \(\mathcal{A}\) and \(j_{\mathbb{T}}\) is the unique functor from \(\mathcal{A}\) to \(\Theta_{\mathbb{T}}\) such that \(i_{\mathbb{T}}\circ j_{\mathbb{T}} = F\circ i_{\mathcal{A}}\), where \(i_{\mathbb{T}}\) is the full inclusion from \(\Theta_{\mathbb{T}}\) to \({\mathcal EM}(\mathbb{T})\)). From the nerve theorem the authors obtain, as a corollary, the classical result of \textit{P. Gabriel} and \textit{F. Ulmer} [Lokal präsentierbare Kategorien. Berlin: Springer (1971; Zbl 0225.18004)] concerning Eilenberg-Moore categories of \(\alpha\)-accessible monads in \(\alpha\)-accessible categories. Section 2 is devoted to alternative formulations of the concept of monad with arities (e.g., they prove in Proposition 2.3 that monads with arities are exactly the monads of the \(2\)-category of categories with arities, arity-respecting functors and natural transformations). Moreover, the authors study strongly Cartesian monads (i.e., Cartesian monads such that the underlying endofunctor is a local right adjoint (\(\equiv\) parametric right adjoint)) and point out that local right adjoints allow a calculus of generic factorizations. I now recall, following the authors, the concept of generic morphism relative to a functor that will be used by them in Theorem 2.9 (see below). Given a functor \(R\) from a category \(\mathcal{E}\) to another \(\mathcal{F}\), a morphism \(g\) from \(B\) to \(R(A)\) is called \(R\)-generic whenever, given morphisms \(\alpha\) from from \(B\) to \(R(A')\), \(\beta\) from \(A\) to \(X\), and \(\gamma\) from \(A'\) to \(X\), if \(R(\beta)\circ g = R(\gamma)\circ \alpha\), then there exists a unique morphism \(\delta\) from \(A\) to \(A'\) such that \(R(\delta)\circ g = \alpha\) and \(\beta = \gamma\circ \delta\). As I have said above this is used by the authors in Theorem 2.9 to show that every strongly Cartesian monad \(\mathbb{T}\) comes equipped with canonical arities \(\mathcal{A}_{\mathbb{T}}\). More precisely, they state in Theorem 2.9, essentially, the following: ``If \(\mathbb{T}\) is a strongly Cartesian monad on a finitely complete category \(\mathcal{E}\), then any dense generator \((\mathcal{A},i_{\mathcal{A}})\) of \(\mathcal{E}\) embeds in a minimal \(T\)-generically closed dense generator \((\mathcal{A}_{\mathbb{T}},i_{\mathcal{A}_{\mathbb{T}}})\)'' (where, I recall, following once more the authors, that a generator \((\mathcal{A},i_{\mathcal{A}})\) of \(\mathcal{E}\) is called \(T\)-generically closed if for any \(T\)-generic morphism from \(B\) to \(T(A)\), with \(B\in \mathcal{A}\), there is an object isomorphic to \(A\) which belongs to \(\mathcal{A}\)). Moreover, the authors point out that the free \(\omega\)-category monad \(\mathbb{D}_{\omega}\) on the category of globular sets is a strongly Cartesian monad so that the monads induced by \(\omega\)-operads have canonical arities. Because it will be used below, when explaining the contents of the following section, I recall that, for a functor \(j\) from \(\mathcal{A}\) to \(\mathcal{B}\), the left and right adjoints to the restriction functor \(j^{\ast}\) from \(\widehat{\mathcal{B}}\) to \(\widehat{\mathcal{A}}\) are denoted by the authors by \(j_{!}\) and \(j_{\ast}\) respectively. Section 3 introduces, for a category \(\mathcal{E}\) with dense generator \((\mathcal{A},i_{\mathcal{A}})\), (1) the concept of theory with arities \((\mathcal{A},i_{\mathcal{A}})\) on \(\mathcal{E}\) as a pair \((\Theta,j)\), where \(\Theta\) is a category and \(j\) a bijective-on-objects functor from \(\mathcal{A}\) to \(\Theta\) such that the induced monad \(j^{\ast}\circ j_{!}\) on \(\widehat{\mathcal{A}}\) preserves the essential image of the nerve functor \(\nu_{\mathcal{A}}\) from \(\mathcal{E}\) to \(\widehat{\mathcal{A}}\), (2) for two theories \((\Theta_{0},j_{0})\) and \((\Theta_{1},j_{1})\) (with arities \((\mathcal{A},i_{\mathcal{A}})\) on \(\mathcal{E}\)), the concept of morphism of theories from \((\Theta_{0},j_{0})\) to \((\Theta_{1},j_{1})\) as a functor \(\theta\) from \(\Theta_{0}\) to \(\Theta_{1}\) such that \(j_{1} = \theta\circ j_{0}\), (3) the concept of \((\Theta,j)\)-model as a presheaf on \(\Theta\) whose restriction along \(j\) belongs to to the essential image of \(\nu_{\mathcal{A}}\), and (4) a morphism of \((\Theta,j)\)-models as a natural transformation of the underlying presheaves. The category of theories with arities \((\mathcal{A},i_{\mathcal{A}})\) on \(\mathcal{E}\) is denoted by the authors as \({\mathcal Th}(\mathcal{E},\mathcal{A})\) and the category of \((\Theta,j)\)-models as \({\mathcal Mod}_{\Theta}\) (which is a full subcategory of \(\widehat{\Theta}\)). Then the equivalence between monads and theories for a fixed category with arities is established in Theorem 3.4 as follows: ``Let \(\mathcal{E}\) be a category with dense generator \((\mathcal{A},i_{\mathcal{A}})\). Then there exists an adjoint equivalence between \({\mathcal Mnd}(\mathcal{E},\mathcal{A})\) and \({\mathcal Th}(\mathcal{E},\mathcal{A})\)'' [defined, at the level of objects, by associating to \(\mathbb{T}\) in \({\mathcal Mnd}(\mathcal{E},\mathcal{A})\) precisely \((\Theta_{\mathbb{T}},j_{\mathbb{T}})\) in \({\mathcal Th}(\mathcal{E},\mathcal{A})\), where \(\Theta_{\mathbb{T}}\) is the full subcategory of \({\mathcal EM}(\mathbb{T})\) spanned by the free \(\mathbb{T}\)-algebras on the objects of \(\mathcal{A}\) and \(j_{\mathbb{T}}\) the (restricted) free \(\mathbb{T}\)-algebra functor]. From this the authors obtain as a particular case the equivalence between finitary monads on the category \({\mathcal Set}\) of sets and Lawvere's algebraic theories. Following this they define the concept of homogeneous theory (as those theories \((\Theta,j)\) with arities \((\mathcal{A},i_{\mathcal{A}})\) on \(\mathcal{E}\) which contain \(\mathcal{A}\) as a subcategory and which admit a generic/free factorization system), and obtain in Theorem 3.10, for each strongly Cartesian monad \(\mathbb{T}\) (whose arities have no symmetries) a correspondence between \(\mathbb{T}\)-operads and \(\Theta_{\mathbb{T}}\)-homogeneous theories. This yields, in particular, the canonical equivalence between the category of Batanin's globular \(\omega\)-operads and the category of \((\Theta_{\omega},j_{\omega})\)-homogeneous theories, where \((\Theta_{\omega},j_{\omega})\) denotes the homogeneous globular theory associated to \(\mathbb{D}_{\omega}\). In 3.14 the authors show that symmetric operads can be considered as \(\Gamma\)-homogeneous theories, where \(\Gamma\) is \textit{G. Segal}'s category [Topology 13, 293--312 (1974; Zbl 0284.55016)]. Finally, Section 4 is devoted to the study of the free groupoid monad on the category of involutive graphs. In Theorem 4.15 the authors show that the finite connected acyclic graphs endow the free groupoid monad with arities, and that this may be used to recover Grothendieck's symmetric simplicial characterization of groupoids [\textit{A. Grothendieck}, in: Techniques de construction et théorèmes d'existence en géométrie algébrique. III: Préschémas quotients'. Séminaire Bourbaki 13, No. 212. Paris: Société Mathématique de France (1961; Zbl 0235.14007)] as an instance of the nerve theorem.