When can a formality quasi-isomorphism over \(\mathbb{Q}\) be constructed recursively? (Q1713033): Difference between revisions

A differential graded (dg) operad $\mathcal{O}$ is formal if there exists a sequence of quasi-isomorphisms of dg operads connecting $\mathcal{O}$ to its cohomology $H^{\bullet}(\mathcal{O})$. As was pointed by the authors of the paper under review, there are no effective tools for determining whether a given dg operad is formal or not and in all known cases the proofs of formality require transcendental tools. If we assume that $\mathcal{O}$ is a differential graded (possibly colored) operad defined over rationals and we also assume that there exists a zig-zag of quasi-isomorphisms connecting $\mathcal{O}\otimes\mathbb{K}$ to its cohomology, where $\mathbb{K}$ is any field extension of $\mathbb{Q}$, in this paper the authors show that for a large class of such dg operads, a formality quasi-isomorphism for $\mathcal{O}$ exists and can be constructed recursively. More concretely, let $\mathcal{O}$ be a dg operad, let $\mathcal{H}=:=H^{\bullet}(\mathcal{O})$ be the cohomology operad for $\mathcal{O}$ and assume that $\mathcal{H}$ admits a cobar resolution $\mathrm{Cobar} (P^{\diamond})$ where $P^{\diamond}$ is a dg pseudo-cooperad satisfying that there exists a quasi-isomorphism of dg operads $\mathrm{Cobar} (P^{\diamond})\stackrel{\sim}{\longrightarrow}\mathcal{O}$, and $P^{\diamond}$ is the dg cooperad which is obtained from $P$ by formally adjoining the counit. Assume that $\mathcal{O} \otimes\mathbb{K}$ is formal for some field extension $\mathbb{K}$ of $\mathbb{Q}$, let $n$ be an integer, $n\ge 2$, such that $ \alpha = \alpha_1 + \dots + \alpha_n \in \mathrm{Conv}(P, \mathcal{O}), (\alpha_k \in \mathrm{Hom}_{\mathsf{Coll}}(\mathcal{G}^k P, \mathcal{O} ))$ is an $n$-th MC-sprout in $\mathrm{Conv}(P, \mathcal{O} )$ for which the diagram \[ \begin{tikzcd} & \mathcal{Z} (\mathcal{O})\ar[d, "\pi_{\mathcal{H}}"] \\ \mathcal{G}^1 P\ar[r, "\alpha_{\mathcal{H}}"] \ar[ur, "\alpha"] & \mathcal{H} \end{tikzcd} \] commutes, and finally suppose that the pair $(P, \mathcal{O} )$ satisfies the conditions: C1: The dg pseudo-operad $P$ is equipped with an additional grading \[ P = \bigoplus_{k \geq 1} \mathcal{G}^k P\,, \qquad \mathcal{G}^{\,\le 0} P = \mathbf{0} \] which is compatible with the differential $\partial_P$ and the comultiplications $\Delta_{\mathbf{t}}$ in the following sense: \[ \partial_P (\mathcal{G}^k P) \subset \mathcal{G}^{k-1} P, \] \[ \Delta_{\mathbf{t}}(\mathcal{G}^m P) \subset \bigoplus_{r_1+..+ r_q = m} \mathcal{G}^{r_1}P \otimes \mathcal{G}^{r_2}P \otimes \dots \otimes \mathcal{G}^{r_q}P, \] where ${\mathbf{t}}$ is ($\Xi$-colored) tree with $q$ nodal vertices, C2: $\mathcal{G}^k P$ is finite dimensional for every $k$ and the graded components of $\mathcal{O}(\mathbf{q})$ are finite dimensional for every $\Xi$-colored corolla $\mathbf{q}$. C3: The operad $\mathcal{H}$ is generated by $\rho (\mathbf{s} \mathcal{G}^1 P)$ and $$ \rho \big|_{\mathbf{s} \mathcal{G}^k P} = 0 \; \forall k \ge 2. $$ Then, under these conditions the authors prove that the quasi-isomorphism of operads $\mathrm{Cobar} (P^{\diamond}) \, \stackrel{\sim}{\longrightarrow} \, \mathcal{O} $ can be constructed recursively, and the algorithm for constructing this quasi-isomorphism involves a solution of a finite dimensional linear system and it requires no explicit knowledge about the quasi-isomorphisms connecting $\mathcal{O} \otimes \mathbb{K}$ to $\mathcal{O} \otimes \mathbb{K}$.