Milnor-Moore categories and monadic decomposition (Q903948)

Let \((L:\mathcal{B}\to\mathcal{A},R:\mathcal{A}\to\mathcal{B})\) be an adjunction with unit \(\eta\) and counit \(\epsilon\). Then \(RL\) is a monad on \(\mathcal{B}\) and one can consider the Eilenberg-Moore category \(_{RL}\mathcal{B}\) associated to this monad and the comparison functor \(K:\mathcal{A}\to_{RL}\mathcal{B}\) defined by \(KX:=(RX,R\epsilon X)\) and \(Kf:=Rf\). In the case when \(K\) itself has a left adjoint \(\Lambda\), one can repeat this construction starting from the new adjunction \((\Lambda, K)\). Changing the notation \((L, R)\) and \((\Lambda, K)\) by \((L_0, R_0)\) and \((L_1, R_1)\) respectively, one obtains a diagram of the form \[ \begin{tikzcd} \mathcal{A} \ar[d, "R_0", shift left=0.5 ex] & \mathcal{A} \ar[l,"\mathrm{Id}_{\mathcal{A}}" '] \ar[d, "R_1", shift left=0.5 ex] &\mathcal{A} \ar[l,"\mathrm{Id}_{\mathcal{A}}" '] \ar[d, "R_2", shift left=0.6 ex] & \dots\ar[l,"\mathrm{Id}_{\mathcal{A}}" '] \\ \mathcal{B}_0 \ar[u, "L_0", shift left=0.5ex] & \mathcal{B}_1 \ar[u, "L_1", shift left=0.5ex]\ar[l, "U_{0,1}"] & \mathcal{B}_2 \ar[u, "L_2", shift left=0.5ex]\ar[l, "U_{1,2}"] & \dots \ar[l, "U_{2,3}"] \end{tikzcd}\tag{1} \] If there is a minimal \(N\in\mathbb{N}\) such that \(L_N\) is full and faithful, then \(R\) is said to have monadic decomposition of monadic length \(N\). This is equivalent to requiring that the forgetful functor \(U_{N, N+1}\) is a category isomorphism and no \(U_{n, n+1}\) has this property for \(0\leq n\leq N-1\). In [\textit{A. Ardizzoni} et al., Appl. Categ. Struct. 23, No. 1, 93--105 (2015; Zbl 1310.18002)] the particular case of diagram (1) where \(\mathcal{A}=\mathrm{Bialg}_{\mathfrak{M}}\) (the category of \(\mathbf k\)-bialgebras) and \(\mathcal{B}_0=\mathfrak{M}\) (the category of vector spaces over a field base \(\mathbf k\)), \(L_0=\overline{T}\) is the tensor bialgebra functor and \(R_0=P\) is the primitive functor which assigns to each \(\mathbf k\)-bialgebra its space of primitive elements, was analyzed and was proved that \(P\) has monadic decomposition of monadic length at most 2. Moreover, when \(\mathrm{char}(\mathbf k)=0\), for every \(V_2=((V, \mu),\mu_1)\in\mathfrak{M}_2\) one can define \([x,y]:=\mu(xy-yx)\) for every \(x,y \in V\). Then \((V,[-,-])\) is a Lie algebra and \(\overline{T}_2V_2=\frac{TV}{(xy-yx-[x,y]\mid x,y\in V)}\) is the corresponding universal enveloping algebra. This fact suggest a connection between the category \(\mathfrak{M}_2\) and the category \(\mathrm{Lie}_{\mathfrak{M}}\) of Lie \(\mathbf k\)-algebras. So it is natural to expect the existence of a category equivalence \(\Lambda\) such that the following diagram is commutative in its undashed part \[ \begin{tikzcd} \mathrm{Bialg}_{\mathfrak{M}} \ar[d,"P", shift left=0.5ex] & \mathrm{Bialg}_{\mathfrak{M}}\ar[d,"P_1", shift left=0.5ex] \ar[l, "\mathrm{Id}_{\mathrm{Bialg}_{\mathfrak{M}}}" '] & \mathrm{Bialg}_{\mathfrak{M}} \ar[d,"P_2", shift left=0.5ex] \ar[l, "\mathrm{Id}_{\mathrm{Bialg}_{\mathfrak{M}}}" '] \\ \mathfrak{M} \ar[u,"\overline{T}", shift left=0.5ex] & \mathfrak{M}_1 \ar[u,"\overline{T}_1", shift left=0.5ex] \ar[l, "U_{0,1}"] & \mathfrak{M}_2 \ar[u,"\overline{T}_2", shift left=0.5ex] \ar[l, "U_{1,2}"]\end{tikzcd} \] where \(H_{\mathrm{Lie}}\) denotes the forgetful functor, \(\overline{U}\) the universal enveloping bialgebra functor and \(P\) the corresponding primitive functor. To solve the above problem it is more natural to work with braided \(\mathbf k\)-vector spaces and to replace the categories \(\mathfrak{M},\mathrm{Bialg}_{\mathfrak{M}}\) and \(\mathrm{Lie}_{\mathfrak{M}}\) by their braided analogues, denoted by \(\mathrm{Br}_{\mathfrak{M}}\), \(\mathrm{BrBialg}_{\mathfrak{M}}\) and \(\mathrm{BrLie}_{\mathfrak{M}}\), respectively. Further \(\mathfrak{M}\) is substituted by an arbitrary monoidal category \(\mathcal{M}\). Moreover, to produce a braided analogue of the universal enveloping algebra which carries a braided bialgebra structure, it is necessary to assume that the underlying Yang-Baxter operator is symmetric. Thus \(\mathrm{Br}^s_{\mathcal{M}}\), \(\mathrm{BrBialg}^s_{\mathcal{M}}\) and \(\mathrm{BrLie}^s_{\mathcal{M}}\) are the analogue of \(\mathrm{Br}_{\mathcal{M}}\), \(\mathrm{BrBialg}_{\mathcal{M}}\) and \(\mathrm{BrLie}_{\mathcal{M}}\) consisting of objects with symmetric Yang-Baxter operator. Let \(\overline{T}^s_{\mathrm{Br}}:\mathrm{Br}^s_{\mathcal{M}}\to \mathrm{BrBialg}^s_{\mathcal{M}}\) be the symmetric braided tensor bialgebra functor and let \(P^s_{\mathrm{Br}}\) be its right adjoint, the primitive functor. Authors look for a condition for \(P^s_{\mathrm{Br}}\) to have monadic decomposition of monadic length at most two. Moreover, \(P^s_{\mathrm{Br}}\) induces a functor \(\mathcal P^s_{\mathrm{Br}}:\mathrm{BrBialg}^s_{\mathcal{M}}\to\mathrm{BrLie}^s_{\mathcal{M}}\) whose left adjoint is the universal enveloping bialgebra functor \(\overline{\mathcal U}^s_{\mathrm{Br}}\). A category \(\mathcal{M}\) is said to be a MM-category (Milnor-Moore category) if the unit of the adjunction \((\overline{\mathcal U}^s_{\mathrm{Br}},\mathcal P^s_{\mathrm{Br}})\) is a functorial isomorphism (plus other required conditions). One of the main results in the paper establishes that the functor \(\mathcal P^s_{\mathrm{Br}}\) has monadic decomposition of monadic length at most two, provided that \(\mathcal{M}\) is a MM-category. Moreover, in this case, the category \((\mathrm{Br}^s\mathcal{M})_2\) can be identified with \(\mathrm{BrLie}^s_{\mathcal{M}}\) through an equivalence \(\Lambda_{\mathrm{Br}}:(\mathrm{Br}^s\mathcal{M})_2\to\mathrm{BrLie}^s_{\mathcal{M}}\). Hence MM-categories give an environment where the functor \(P^s_{\mathrm{Br}}\) has a behavior analogous to the classical vector spaces setting. Following, meaningful examples of MM-categories are provided, thus authors prove that the category of vector spaces over a fixed field \(\mathbf k\) of characteristic zero is a MM-category (Theorem 8.1) and the central result (Theorem 8.4) establishes: For an abelian monoidal category \(\mathcal{M}\) with denumerable coproducts and such that the tensor products are exact and preserve denumerable coproducts, and a MM-category \(\mathcal{N}\), assume that there exists a conservative and exact monoidal functor \((F, \phi_0, \phi_2) : \mathcal{M}\to\mathcal{N}\) which preserves denumerable coproducts, then \(\mathcal{M}\) is a MM-category. Authors use this central result in the case of the forgetful functor \(F:\mathcal{M}\to\mathfrak{M}\), where \(\mathcal{M}\) is a subcategory of \(\mathfrak{M}\). The goal is to provide examples of MM-categories \(\mathcal{M}\) and, in the case when \(\mathcal{M}\) is symmetric, to recognize the corresponding type of Lie algebras. An example of MM-category obtained in this way is the category of Yetter-Drinfield modules, over a Hopf algebra over a field of characteristic zero. When \(\mathcal{M}\) is the category of modules (resp. comodules) over a quasi-bialgebra (reps. the dual quasi-bialgebra), it is proved that the forgetful functor satisfies the central theorem if and only if the quasi-bialgebra (resp. the dual quasi-bialgebra) is a deformation of a usual bialgebra. As particular case of this situation, it is proved that monoidal Hom-Lie algebras, Lie color algebras, Lie superalgebras and other type of generalized Lie algebras are MM-categories. The paper ends with two appendices containing general results on the existence of (co)equalizers in the category of (co)algebras, bialgebras and their braided analogue over a monoidal category.