The aim of this note is to give a detailed account of how symmetric operads can be constructed from planar (non-symmetric) operads, and to carefully spell out the algebraic interplay between these two notions. It is a companion note to the main paper "Homotopical operadic calculus in positive characteristic".
A dg \mathbb N-modules amounts to the data of a functor
The object X(n) is called the arity n part of X. We denote \mathsf{dg}~\mathbb N\text{-}\mathsf{mod} the category of dg \mathbb N-modules.
A planar dg operad \mathcal{P} amounts to the data of a monoid (\mathcal{P},\gamma,\eta) in the category of dg \mathbb N-modules with respect to the composition product.
An augmented planar dg operad \mathcal{P} amounts to the data of a planar dg operad (\mathcal{P},\gamma,\eta) equipped with a morphism of planar dg operads \nu: \mathcal{P} \longrightarrow \mathcal{I} such that \nu \circ \eta = \mathrm{id}.
A planar dg cooperad \mathcal{C} amounts to the data of a comonoid (\mathcal{C}, \Delta, \epsilon) in the category of dg \mathbb N-modules with respect to the composition product.
A coaugmented planar dg cooperad \mathcal{C} amounts to the data of a planar dg cooperad (\mathcal{C}, \Delta, \epsilon) equipped with a morphism of planar dg cooperads \mu: \mathcal{I} \longrightarrow \mathcal{C} such that \epsilon \circ \mu = \mathrm{id}.
One can combine these notations. For instance the n-levelled reduced planar tree endofunctor \overline{\mathbb{T}}_{\mathrm{pl},\leq n}(X) is obtained by taking the sum over non trivial trees whose height is at most n.
Notice that, for all n \geq 0, there is a canonical isomorphism
that yields an isomorphism \mathbb{T}_\mathrm{pl} X \cong \mathcal{I} \oplus X \circ \mathbb{T}_{\mathrm{pl}} X.
The planar tree module endofunctor \mathbb{T}_\mathrm{pl} admits the following monad structure:
The unit \eta_X: X \longrightarrow \mathbb{T}_\mathrm{pl} X is the canonical inclusion, where elements in X(m) are seen as m-corollas.
The product \gamma_X: \mathbb{T}_\mathrm{pl}~\mathbb{T}_\mathrm{pl}(X) \longrightarrow \mathbb{T}_\mathrm{pl} (X) is given, for a planar tree t with n leaves, as follows
where \mathrm{graft}_t(t_1, \ldots, t_k) is the planar tree obtained from t by replacing each node i by the planar tree t_i which labelled it.
The category of algebras over the monad \mathbb{T}_\mathrm{pl} is canonically isomorphic to the category of planar dg operads.
The restriction of the structural map to \mathbb{T}_{\mathrm{pl}, \leq 2} gives the composition product, and induces an operad structure. Any operad structure can be extended to a full algebra structure over \mathbb{T}_\mathrm{pl} by induction on the level of planar trees.
The reduced planar tree module endofunctor \overline{\mathbb{T}}_\mathrm{pl} admits the following comonad structure:
The counit map \epsilon_X: \overline{\mathbb{T}}_\mathrm{pl}(X) \longrightarrow X is the projection onto corollas labelled by X.
The coproduct \Delta_X: \overline{\mathbb{T}}_\mathrm{pl}(X) \longrightarrow \overline{\mathbb{T}}_\mathrm{pl} ~\overline{\mathbb{T}}_\mathrm{pl}(X) is given by partitioning non-trivial planar trees into non-trivial planar sub-trees. For a planar tree t with n leaves, the coproduct is given by the finite sum over partitions t_1, \ldots, t_m of t into non trivial sub-trees of the maps
This follows from a straightforward checking.
For every natural integer n \geq 1, the comonad structure on \overline{\mathbb{T}}_\mathrm{pl} restricts to \overline{\mathbb{T}}_{\mathrm{pl},\leq n} and \overline{\mathbb{T}}_\mathrm{pl}^{(\leq n)}, making the natural inclusions
morphisms of comonads.
This follows from the definition of the coproduct of the comonad. For every non-trivial tree t with at most n-nodes (resp. of height at most n) and every partition t_1, \ldots, t_m of t, all the planar trees t_1, \ldots, t_m, t/(t_1, \ldots, t_m) have at most n-nodes (resp. have height at most n).
For every dg \overline{\mathbb{T}}_\mathrm{pl}-coalgebra (W,\delta_W), the following square
is a commutative diagram.
This square decomposes into the diagram
of which all the cells are commutative.
The functor \mathrm{Conil} from dg \overline{\mathbb{T}}_\mathrm{pl}-coalgebras to coaugmented planar dg cooperads is fully faithful.
This functor is faithful since its composition with the forgetful functor towards dg \mathbb N-modules is faithful. Let us prove that it is is full. Let (V,\delta_V) and (W,\delta_W) be two dg \overline{\mathbb{T}}_\mathrm{pl}-coalgebras and let us consider a morphism of coaugmented planar dg cooperads f: \mathcal{I} \oplus V \to \mathcal{I} \oplus W. Proving that f stems from a morphism of dg \overline{\mathbb{T}}_\mathrm{pl}-coalgebras amounts to prove that the restricted map \overline{f}: V \to W is a morphism of dg \overline{\mathbb{T}}_\mathrm{pl}-coalgebras.
Since the natural map \overline{\mathbb{T}}_\mathrm{pl} X \to \lim_{n \in \omega^\mathrm{op}}\overline{\mathbb{T}}_{\mathrm{pl},\leq n} X is a monomorphism, it suffices to prove that for every natural integer n \geq 1 the diagram
is commutative.
For n=1, this square is clearly commutative. Let us assume that it is commutative for an integer n \geq 1 and let us prove that is is commutative for n+1. The square above for n+1 rewrites as the external square of the diagram
of which all cells are commutative. So we have proven by induction that such a square is commutative for every natural integer n \geq 1. Therefore \overline{f} is a morphism of dg \overline{\mathbb{T}}_\mathrm{pl}-coalgebras.
Let \mathcal{C} be a coaugmented planar dg cooperad. It is conilpotent if it belongs to the essential image of the functor \mathrm{Conil} from dg \overline{\mathbb{T}}_\mathrm{pl}-coalgebras to planar dg cooperads. We denote \mathsf{dg}~\mathsf{Cooperads}^{\mathsf{conil}}_\mathrm{pl} the full sub-category of coaugmented planar dg cooperads spanned by conilpotent ones.
The idea behind this definition is the following: a cooperad can also be described in terms of partial decomposition maps
and it is conilpotent if and only if any iteration of these partial decompositions is eventually trivial. If this is the case, then the data of all the possible iterations is exactly encoded by the \overline{\mathbb{T}}_\mathrm{pl}-coalgebra structure.
Let n be a natural integer. The forgetful functor
is fully faithful. Furthermore, both categories are presentable, and therefore it admits a right adjoint denoted by \overline{\mathrm{F}}_n.
The two categories are presentable. It follows from the fact that they are categories of coalgebras over accessible comonads on a presentable category. We refer to [ChingRiehl] for more details. The fact that the functor is fully faithful follows from the fact that the morphism of comonads \overline{\mathbb{T}}_{\mathrm{pl}}^{(\leq n)} \longrightarrow \overline{\mathbb{T}}_{\mathrm{pl}} is object-wise a monomorphism. The fact that it has a right adjoint is a direct consequence of the adjoint lifting theorem, see [premierpapier, Appendix A].
Let \mathcal{C} be a conilpotent planar dg cooperad. Its n-coradical filtration is given by the conilpotent planar dg cooperad \mathrm{F}^{\mathrm{rad}}_n \mathcal{C}. It induces a ladder diagram
indexed by \mathbb{N}, where all the arrows are monomorphisms.
Let n \geq 0 and \mathcal{C} be a conilpotent planar dg cooperad. Then \mathrm{F}^{\mathrm{rad}}_n \mathcal{C} fits in the following pullback
in the category of dg \mathbb{N}-modules, where \delta_\mathcal{C} denotes the dg \overline{\mathbb{T}}_{\mathrm{pl}}-coalgebra structure of \mathcal{C}.
This directly follows from the results in [premierpapier, Appendix A].
Let \mathbb S be the groupoid whose objects are natural integers and whose morphisms are given by
A dg \mathbb S-module M amounts to the data of a functor
from \mathbb S^\mathrm{op} to dg modules. We denote by \mathsf{dg}~\mathbb{S}\text{-}\mathsf{mod} the category of dg \mathbb{S}-modules.
The underlying \mathbb N-module of u\mathcal{A} ss has the structure of a planar operad.
The unit is given by the identity map of \Bbbk.
The composition is defined as follows. For every (\sigma, \sigma_1, \ldots, \sigma_k) \in \mathbb S_k \times \mathbb S_{n_1} \times \cdots \times \mathbb S_{n_k} the composition of \sigma \otimes \sigma_1 \otimes \cdots \otimes \sigma_k is given by the permutation
in u\mathcal{A} ss (n), where n = n_1 + \cdots + n_k.
This follows from a straightforward check.
The endofunctor -\otimes \mathbb S of the category of dg \mathbb N-modules has a canonical structure of lax monoidal functor with respect to the planar composition product \circ_{\mathrm{pl}}.
There is a canonical isomorphism \mathcal{I} \cong \mathcal{I} \otimes \mathbb S given by the identity map,
and a binatural map of dg \mathbb{N}-modules
given by
where \lambda is the composition in u\mathcal{A} ss of \sigma_k \otimes (\sigma_{n_1} \otimes \cdots \otimes \sigma_{n_k}).
This amounts to showing that the following diagrams
and
are commutative for every dg \mathbb N-modules X,Y,Z. This follows from the associativity and the unitality of the compositions in the planar operad u\mathcal{A} ss.
There are natural transformations
and
of endofunctors in the category of dg \mathbb N-modules, induced, respectively, by the units and the multiplications in the group algebras \Bbbk[\mathbb{S}_n] for all n \geq 0. Furthermore, these natural transformations are monoidal natural transformations between lax monoidal functors.
The behaviour of the monad - \otimes \mathbb S with respect to the unit \mathcal{I} is straightforward to check. We check that the two following diagrams
and
are commutative for all dg \mathbb{N}-modules X and Y. It follows from the compatibility of the compositions in the planar operad u\mathcal{A} ss with the actions of the symmetric groups.
There is a natural transformation
where \mathrm{U} is the forgetful functor from dg \mathbb{S}-modules to dg \mathbb{N}-modules. For M a dg \mathbb S-modules, \pi_M: \mathrm{U}(M) \otimes \mathbb{S} \longrightarrow M is given by the dg \mathbb{S}-module structure of M.
Let M, N be two dg \mathbb S-modules. We define the composition product M \circ N as the coequaliser of the reflexive pair of maps
in the category of dg \mathbb S-modules.
For every two dg \mathbb N-modules X,Y, the natural map
is an isomorphism of dg \mathbb S-modules.
Let us consider the composite map
Using Lemma 3 in a straightforward way, one can prove that it factors through (X \otimes \mathbb S) \circ (Y\otimes \mathbb S) and that the composite map
is the identity. To prove that the composite map
is the identity, it is enough to prove that the composition
given by \psi_{X,Y}~\mu_{X \circ_\mathrm{pl} Y}~(\varphi_{X,Y} \otimes \mathbb S)~\rho_{X \otimes \mathbb S ,Y \otimes \mathbb S} is equal to the canonical quotient map \rho_{X \otimes \mathbb S ,Y \otimes \mathbb S}. Using Lemma 3 again, we can rewrite this composition as:
The universal property that defines the composition product - \circ - allows us to change the second map \mu_{(X \otimes \mathbb{S})\circ_\mathrm{pl} (Y \otimes \mathbb{S})} ~(\varphi_{X \otimes \mathbb{S},Y \otimes \mathbb{S}} \otimes \mathbb{S}) by (\mu_X \circ_\mathrm{pl} \mu_Y ) \otimes \mathbb{S} without modifying the whole composition. Then, the composition of the two first arrows is the identity, which implies that the whole composition is equal to its last component \rho_{X \otimes \mathbb S ,Y \otimes \mathbb S}, which concludes the proof.
Let M, N be two dg \mathbb S-modules. There is a natural isomorphism
of dg \mathbb S-modules, where \mathrm{Ind} stands for the induced representation.
By Proposition 8, both expressions are isomorphic for free dg \mathbb{S}-modules. Furthermore, both expressions commute with sifted colimits. Hence both expressions are isomorphic for any pair of dg \mathbb{S}-modules.
Therefore Definition 9 coincides with the standard definition of the composition product of dg \mathbb{S}-modules present in the literature. See for instance [LodayVallette].
The composition product \circ defines the structure of a monoidal category on dg \mathbb S-module whose unit is given by \mathcal{I}.
Let X,Y,Z be three dg \mathbb{N}-modules. One can construct an associator isomorphism
using the binatural isomorphism \psi of Proposition 8 and the associator isomorphism of the composition product \circ_\mathrm{pl} in the category of dg \mathbb{N}-modules. Furthermore, this associator isomorphism commutes with the natural transformation \mu: - \otimes \mathbb S \otimes \mathbb S \longrightarrow - \otimes \mathbb S, hence it induces an associator isomorphism for all free dg \mathbb{S}-modules. Using the compatibility of these constructions with sifted colimits, it induces a associator isomorphism for any three dg \mathbb{S}-modules. Likewise for the unitor isomorphisms. The commutativity of the required diagrams then follows from the commutativity of the unitors and the associators of the planar composition product - \circ_\mathrm{pl} -.
The functor -\otimes \mathbb S from dg \mathbb N-modules to dg \mathbb S-modules is strong monoidal.
This structure of a strong monoidal functor is given by the identity map \mathcal{I} \longrightarrow \mathcal{I} \otimes \mathbb S and the binatural map \psi of Proposition 8. The fact that these maps define the structure of a strong monoidal functor is a direct consequence of the fact that they are used in the definition of the associator and the unitors related to the composition product.
A dg operad \mathcal{P} amounts to the data of a monoid (\mathcal{P},\gamma,\eta) in the category of dg \mathbb N-modules with respect to the composition product.
An augmented dg operad \mathcal{P} amounts to the data of a dg operad (\mathcal{P},\gamma,\eta) equipped with a morphism of dg operads \nu: \mathcal{P} \longrightarrow \mathcal{I} such that \nu \circ \eta = \mathrm{id}.
A dg cooperad \mathcal{C} amounts to the data of a comonoid (\mathcal{C}, \Delta, \epsilon) in the category of dg \mathbb S-modules with respect to the composition product.
A coaugmented dg cooperad \mathcal{C} amounts to the data of a dg cooperad (\mathcal{C}, \Delta, \epsilon) equipped together with a morphism of planar dg cooperads \mu: \mathcal{I} \longrightarrow \mathcal{C} such that \epsilon \circ \mu = \mathrm{id}.
The adjunction - \otimes \mathbb S \dashv U_{\mathbb S} relating dg operads to planar dg operads is monadic since its right adjoint preserves coreflexive equalisers and is conservative. However the other adjunction U_{\mathbb S} \dashv - \otimes \mathbb S \dashv is not a priori comonadic.
The tree intertwining map \tau is the natural morphism of dg \mathbb N-modules
whose first part is given by the units of the monads - \otimes \mathbb S and \mathbb{T}_\mathrm{pl} and whose second part is the monad product of \mathbb{T}_\mathrm{pl}(-) \otimes \mathbb S.
Let g denote a full sub-groupoid of the groupoid of trees \mathsf{Trees}. Let us assume that g is connected within \mathsf{Trees} in the sense that, for any isomorphism of trees t \longrightarrow t', t' belongs to g whenever t does. The planar endofunctor g_\mathrm{pl}(-) associated to g is the endofunctor of dg \mathbb N-modules given, for a dg \mathbb{N}-module X, by
where the sum is taken over planar trees in g up to planar isomorphisms.
For every connected full sub-groupoid g of \mathsf{Trees}, the tree intertwining map
restricts to a morphism
of dg \mathbb{N}-modules.
Let X be a dg \mathbb{N}-module. It is suffices to prove that for every planar tree t, the map
factors through g_\mathrm{pl} (-) \otimes \mathbb S where g is the connected component of the groupoid of trees that contains t. This can be done by induction on the height of trees. On the trivial tree, the tree intertwining map is given by the identity map of \mathcal{I} \otimes \mathbb S. On trees of height one, the tree intertwining map is given by the identity of X \otimes \mathbb S.
Let us assume that it is satisfied for all tree of height equal or lower than m for some natural integer m \geq 1. Let t be a planar tree with n leaves that is made up of a root node with k \geq 1 inputs and topped by k subtrees t_1, \ldots, t_k of height equal or lower than m and with n_1,\ldots, n_k leaves. The height of t is thus lower or equal to m+1. Let us denote g_i the connected component within the groupoid of tree of t_i for i \in \{1, \ldots, k\} and g the connected component of t. On t, the intertwining map is the composition
Here the third map \xi is given by
where \lambda \in \Bbbk[\mathbb S_n] is the composition \sigma \circ (\sigma_1 \otimes \cdots \otimes \sigma_k) in the operad u\mathcal{A} ss. Moreover, for every planar trees t'_1\in g_1, \ldots, t'_k \in g_k, \mathrm{graft}_k(t'_{\sigma^{-1}(1)} ,\ldots, t'_{\sigma^{-1}(k)}) is the planar tree with a root node with k inputs and toppped by the sub-trees, in order, t'_1, \ldots, t'_k. Therefore it is clear that it is isomorphic to t. Thus the map t(X \otimes \mathbb S) \hookrightarrow \mathbb{T}_\mathrm{pl}(X \otimes \mathbb S) \longrightarrow \mathbb{T}_\mathrm{pl}(X) \otimes \mathbb S factors through g_\mathrm{pl}(X) \otimes \mathbb S and we can conclude by induction.
We define the tree monad \mathbb{T} to be the left adjoint of the forgetful functor \mathrm{U}_{\mathrm{Op}}, fitting in the following adjunction
which is furthermore monadic.
Let X be a dg \mathbb{N}-module. The canonical map
is a isomorphism of dg operads, natural in X.
Follows from the definitions.
Let g be a connected full sub-groupoid of the groupoid of trees \mathsf{Trees}. The endofunctor g(-) associated to g is the endofunctor of the category of dg \mathbb{S}-modules given, for a dg \mathbb{S}-module M, by the following coequalizer
where one map is given by the dg \mathbb{S}-module structure of M and the other by the monad structure of - \otimes \mathbb S.
Let M be a dg \mathbb{S}-module, the canonical map
is an isomorphism when the sum is over every connected components of the groupoid \mathsf{Trees}.
Let n be a natural integer. For some particular sub-groupoids g, we use the following notations.
For g being the sub-groupoid of non trivial trees, we denote g(X) as \overline{\mathbb{T}}(X), and call it the reduced tree endofunctor.
For g the sub-groupoid of trees of height at most n, we denote g(X) as \mathbb{T}_{\leq n}(X).
For g the sub-groupoid of trees with n nodes (resp. at most n nodes), we denote g(X) as \mathbb{T}^{(n)}(X).
For g the sub-groupoid of trees with at most n nodes, we denote g(X) as \mathbb{T}^{(\leq n)}(X).
Let X be a dg \mathbb{N}-module. The natural isomorphism
restrict to a natural isomorphism
for every connected full sub-groupoid g of the groupoid of trees \mathsf{Trees}.
It suffices to notice that if \overline{g} is the complement sub-groupoid of g in \mathsf{Trees}, then the map
rewrites as the sum
Since \nu_X is an isomorphism, so are \nu_X|_g and \nu_X|_{\overline g}.
Let g be a connected full sub-groupoid of of the groupoid of trees \mathsf{Trees}. Let M be a dg \mathbb{S}-module. The isomorphism
send the natural diagram in dg \mathbb S-modules
to the natural diagram
whose the maps are given by the monad structure on - \otimes S and the action of this monad on M.
It suffices to prove the result for g the whole groupoid of trees \mathsf{Trees}. This amounts to prove that the following diagrams
are commutative in the category of dg operads. Here the top horizontal map of the left square diagram \mathbb{T}(\mu_{\mathrm{U}(M)}) is induced by the monad product \mu: - \otimes \mathbb S \otimes \mathbb S \longrightarrow -\otimes \mathbb S and the bottom horizontal map is the tree intertwining map \tau. This commutativity follows from the fact that \mathbb{T}(-) is the free dg operad functor.
Let n be a natural integer and M a dg \mathbb{S}-module. The composition of the free operad \gamma_M: \mathbb{T} \circ \mathbb{T} M \longrightarrow \mathbb{T} M induces the following isomorphisms
of dg \mathbb{S}-modules.
First, let us prove the fact that the composition restricts to the isomorphism M \circ \mathbb{T} M \cong \overline{\mathbb{T}} M. The fact that it restricts to the other isomorphisms M \circ \mathbb{T}_{\leq n} M \cong \overline{\mathbb{T}}_{\leq n+1} M will follow from analogue arguments.
When M is a free dg \mathbb S-module of the form X \otimes \mathbb S, this composition rewrites as
which restricts to an isomorphism
The general case follows from the identification of the subsequent coequalizers that define each of the terms.
Let X be a dg \mathbb{N}-module. The tree intertwining map \tau is compatible with the planar cooperad structure of the reduced planar tree comonad. That is, the following diagram
commutes in the category of dg \mathbb{N}-modules. Here \Delta is the planar cooperad structure of \overline{\mathbb{T}}_\mathrm{pl}(X \otimes \mathbb S) the horizontal map \varphi is the lax monoidal functor structure on - \otimes \mathbb S.
Let us prove this result by induction on the height of the trees. This amounts to show that for every natural integer n \geq 1, the pre-compositions by the inclusion \overline{\mathbb{T}}_{\mathrm{pl}, \leq n}(X \otimes \mathbb S) \hookrightarrow\overline{\mathbb{T}}_{\mathrm{pl}}(X \otimes \mathbb S) of the arrows in the square diagram are equal.
For trees of height n=1, the diagram above restricts to
which commutes. Let us assume that the result is satisfied up to a natural integer n \geq 1. Now, let us consider the following diagram
where the top horizontal and the bottom horizontal isomorphisms are,respectively:
where the first one is given by Corollary 2. One can check that both of these isomorphisms are compatible with the planar cooperad structures, hence the top and the bottom squares commute. The second cell square commutes by induction hypothesis and the third by naturality of \varphi. Combined with the fact that the two following square diagrams
commute, it yields the result at level n+1.
Let X be a dg \mathbb{N}-module. The tree intertwining map \tau is compatible with the comonad structure \delta on the reduced tree comonad. That is, the following natural diagrams are commutative
in the category of dg \mathbb{N}-modules.
The fact that the second diagram is commutative is clear. Let us prove that the first one is commutative. As in the proof of Lemma 5, we prove the result by an induction on the height of the trees. More precisely, it suffices to prove that the pre-compositions by the inclusion \overline{\mathbb{T}}_{\mathrm{pl}, \leq n}(X \otimes \mathbb S) \hookrightarrow\overline{\mathbb{T}}_{\mathrm{pl}}(X \otimes \mathbb S) of the arrows in the square diagram are equal. For n=1, this result is clear.
Let us assume that it is satisfied for up to a natural integer n \geq 1. Then, it is satisfied for n+1 since the following diagram is commutative
which is a consequence of Lemma 5 and of the induction hypothesis.
Let X be a dg \mathbb{N}-module. The monad structure \mu of the functor -\otimes \mathbb{S} is compatible with the comonad structure \delta on the reduced tree comonad and the tree intertwining map \tau. That is, the following diagram is commutative
in the category of dg \mathbb{N}-modules.
The commutation of the left side square is a consequence of Proposition 13. The commutation of the middle square is a consequence of Lemma 6. The right square decomposes into the following diagram
which is commutative. Indeed, its top left cell is clearly commutative, its bottom left cell is the left square in the lemma applied to \overline{\mathbb{T}}_\mathrm{pl}(X), and its right cell is the image by \overline{\mathbb{T}}(-) of the same left square in the lemma.
The reduced tree endofunctor \overline{\mathbb{T}} on the category of dg \mathbb S-module has the canonical structure of a comonad. For a dg \mathbb{S}-module M,
its counit \epsilon: \overline{\mathbb{T}}(M) \longrightarrow M is the canonical projection into the corollas,
its coproduct \delta_M: \overline{\mathbb{T}}(M) \longrightarrow \overline{\mathbb{T}}\overline{\mathbb{T}}(M) is the unique arrow which makes the following diagram
commute, where \pi_M: \mathrm{U}(M) \otimes \mathbb{S} \longrightarrow M is the map given by the dg \mathbb{S}-module structure.
The natural map
induces a morphism of diagrams
as a consequence of Lemma 7. Taking the colimits of these diagrams, one obtains a natural coproduct map
It remains to show that this coproduct map and the counit endow the endofunctor \overline{\mathbb{T}}(-) with a comonad structure.
The counitality is straightforward to check. Let us prove that the coproduct map is coassociative. Since \overline{\mathbb{T}}(-) preserves epimorphisms and since the map \mathrm{U}(M) \otimes \mathbb S \longrightarrow M is an epimorphism, it suffices to prove that the following square diagram
is commutative. This follows from the coassociativity of coproduct map \delta_\mathrm{pl}: \overline{\mathbb{T}}_\mathrm{pl} \longrightarrow \overline{\mathbb{T}}_\mathrm{pl}\overline{\mathbb{T}}_\mathrm{pl} and from the fact that the natural isomorphism
is suitably compatible with the coproducts.
One can notice that the natural isomorphism
between functors from the category of dg \mathbb{N}-modules to the category of dg \mathbb S-modules is a comonad transformation above the functor - \otimes \mathbb S. More precisely, this amounts to the following diagrams commuting
This induces by (co)doctrinal adjunction the following corollary.
The adjunction
lifts canonically to an adjunction
Let (X,\delta_X) be a dg \overline{\mathbb{T}}_\mathrm{pl}-coalgebra. The dg \mathbb{S}-module inherits a dg \overline{\mathbb{T}}-coalgebra structure given by
Let (M,\delta_M) be a dg \overline{\mathbb{T}}-coalgebra. The dg \mathbb{N}-module \mathrm{U}(M) inherits a dg \overline{\mathbb{T}}_\mathrm{pl}-coalgebra given by
One can check that these two natural constructions produce adjoint functors.
For every dg \overline{\mathbb{T}}-coalgebra (W,\delta_W), the following square
is a commutative diagram.
The functor \mathrm{Conil} from dg \overline{\mathbb{T}}-coalgebras to coaugmented dg cooperads is fully faithful.
This follows from the same arguments, mutatis mutandis, as those used in the planar context. See the proof of Proposition 4.
Let \mathcal{C} be a coaugmented dg cooperad. It is conilpotent if it belongs to the essential image of the functor \mathrm{Conil} from dg \overline{\mathbb{T}}-coalgebras to dg cooperads. We denote \mathsf{dg}~\mathsf{Cooperads}^{\mathsf{conil}} the full sub-category of coaugmented dg cooperads spanned by conilpotent ones.
The adjunction
restricts to an adjunction
between conilpotent planar dg cooperads and conilpotent dg cooperads.
This follows from the definitions.
Let n be a natural integer.
The comonad structure on the reduced tree endofunctor \overline{\mathbb{T}} restricts to \overline{\mathbb{T}}_{\leq n} and \overline{\mathbb{T}}^{(\leq n)}.
The adjunction
restricts to adjunctions
This follows from the definition of the coproduct of the comonad \overline{\mathbb{T}}. See Corollary 1.
Let n be a natural integer. The forgetful functor
is fully faithful. Furthermore, both categories are presentable, and therefore it admits a right adjoint denoted by \overline{\mathrm{F}}_n.
This follows from the same arguments as in the planar case, see Proposition 5.
Let \mathcal{C} be a conilpotent dg cooperad. Its n-coradical filtration is given by the conilpotent dg cooperad \mathrm{F}^{\mathrm{rad}}_n \mathcal{C}. It induces a ladder diagram
indexed by \mathbb{N}, where all the arrows are monomorphisms.
If the characteristic of the base field \Bbbk is zero, both the tree module \mathbb{T} (-) and the composition product \circ preserve finite cosifted limits. Therefore the forgetful functors from (conilpotent) dg cooperads to dg \mathbb S-modules preserve these limits. Moreover, as in the planar case, given a dg cooperad \mathcal{C}, the conilpotent dg cooperad \overline{\mathrm{F}}_n \mathcal{C} fits in the following pullback diagram
in the category of dg \mathbb{S}-modules, where \delta_\mathcal{C} denotes the dg \overline{\mathbb{T}}-coalgebra structure of \mathcal{C}.
A planar tree t is a tuple (\mathrm{edges}(t), \mathrm{leaves}(t), <, \tilde <) where
(\mathrm{edges}(t), <) is a finite non-empty poset called the poset of edges of t. It has a least element called the root. Moreover, for every edge e, the subposet spanned by elements that are lower that e \{e' \in \mathrm{edges}(t)|\ e' \leq e\} is a linear poset;
\mathrm{leaves}(t) is a subset of the set of maximal elements of the set of edges;
\tilde < is an additional order on the set of edges, called the planar order, that is linear and so that
for every two edges e,e', e ~ \tilde < ~ e' whenever e < e';
for every three edges e,e',e'' so that e < e' and e'' \tilde < e', then e and e'' are comparable for the order < in the sense that either e'' < e or e'' = e or e'' > e.
Let t be a planar tree. A node (or vertex) of t is an edge e that is not a leaf. The inputs of this nodes are the edges e' that are strictly greater than e so that there are no edge between e and e'. Sometimes, the edge that defines the node is called the output edge of this node. Moreover, the number of inputs of a node is called its arity.
Let t,t' be two trees. An iisomorphism of trees from t to t' is an isomorphism of posets from (\mathrm{edges}(t), <) to (\mathrm{edges}(t'), <) that restricts to a bijection between leaves; thus it restricts to a bijection between nodes.
An isomorphism of planar trees from t to t' is an isomorphism of trees that preserves also the planar order.
The groupoid of planar trees \mathsf{Trees}_\mathrm{pl} is the groupoid whose objects are planar trees and whose morphisms are isomorphisms of planar trees.
The groupoid of trees \mathsf{Trees} is the groupoid whose objects are planar trees and whose morphisms are isomorphisms of trees.
A tree t is trivial if \mathrm{edges}(t) = \mathrm{leaves}(t) \cong \ast.
Let t be a planar tree. A subtree t' of t is the data of a tuple (\mathrm{edges}(t'), \mathrm{leaves}(t'), <, \tilde <) where
\mathrm{edges}(t') is a subset of \mathrm{edges}(t) and <, \tilde < are the orders induced from those of t;
\mathrm{leaves}(t) is a subset of the set of maximal elements of the poset of edges of (t', <);
this two subsets satisfy the following conditions
the poset (\mathrm{edges}(t'), <) has a least element (the root of t');
for every three edges e < e' < e'' of t, e' belongs to t' whenever e and e'' do;
for every node n of t that belongs to t', either the underlying edge of n is a leaf of t' or t' has in its edges all the inputs of n;
a leaf of t that belongs to the edges of t' is actually a leaf of t' that is, \mathrm{leaves}(t) \cap \mathrm{edges}(t') \subset \mathrm{leaves}(t).
In particular, t' is a planar tree.
Let t be a non-trivial planar tree. A partition of t into non trivial subtrees is the data of a natural integer m \geq 1 together with non trivial subtrees t_1, \ldots, t_m so that
for every node n of t, there exists a unique i so that n is a node of t_i;
for every leaf l of t, there exists a unique i so that l is an edge (and thus a leaf) of t_i.
Let t be a non-trivial planar tree and let t_1, \ldots, t_m be a partition of t. Then
every node edge e of t belongs to at least one and at most two of these subtrees;
the root of t belongs to only one subtree;
the root of a subtree is a node of this subtree and thus a node of t
an node edge of t can be the root of only one subtree.
This is a straightforward consequence of the definition of a partition into non-trivial subtrees.
Let t be a non-trivial planar tree and let t_1, \ldots, t_m be a partition of t. The quotient tree t/(t_1, \ldots, t_m) of t by this partition is the tree whose nodes are the roots of the subtrees, the leaves are those of t and whose orders <, \tilde < are those of t.
Let t be a non-trivial planar subtree with k nodes t_1, \ldots, t_k be non-trivial trees so that the number of leaves of t_i is equal to the number of inputs of the ith node. This gives a canonical isomorphism
Then we define the grafting tree \mathrm{graft}_t(t_1, \ldots, t_k) of the trees t_1, \ldots, t_k along t as the planar tree so that
the set of node edges is \coprod_{i=1}^k \mathrm{nodes}(t_i);
the set of leaves is the subset of \coprod_{i=1}^k \mathrm{leaves}(t_i) made up of elements whose image in \mathrm{edges}(t) is a leaf of t;
for every two edges e,e' of \mathrm{graft}_t(t_1, \ldots, t_k) that proceeds from the two trees t_i,t_j attached to the nodes n_i, n_j of t, e<e' if and only if either
n_i < n_j and e < l where l is the leaf of t_i that corresponds to the unique input x of the node n_i of t so that x \leq n_j;
i=j and e < e' within t_i;
for every two edges e,e' of \mathrm{graft}_t(t_1, \ldots, t_k) that proceeds from the two trees t_i,t_j attached to the nodes n_i, n_j of t, e<e' if and only if either
n_i ~ \tilde < ~ n_j and n_i and n_j are not comparable for <;
n_i < n_j and e ~ \tilde < ~ l where l is the leaf of t_i that corresponds to the unique input x of the node n_i of t so that x \leq n_j;
i=j and e~ \tilde < ~e' within t_i.
Let t be a non-trivial planar tree and let t_1, \ldots, t_m be a partition of t. One has a canonical isomorphism of planar trees
where t'= t/(t_1, \ldots, t_k).
This follows from a straightforward check.
The text on this page is from arXiv:2312.05005v1 by Brice Le Grignou and Victor Roca i Lucio, under the licence CC BY 4.0. The corpus made these changes: it read the items out of the LaTeX source, it expanded the macros of the paper, it put each proof after its statement, and it converted the text to HTML. The formulas are the TeX of the paper, and KaTeX renders them. A formula or a reference can look different from the PDF. The page is published under CC BY 4.0.
The record of each item is in analysis/objects/2312/2312.05005.jsonl.