The equivalence between Feynman transform and Verdier duality (Q2230807)

The equivalence between cyclic operads with dg-duality and Verdier duality for certain sheaves on the moduli space of stable graphs has been established by \textit{A. Lazarev} and \textit{A. A. Voronov} [Adv. Math. 218, No. 6, 1878--1894 (2008; Zbl 1158.55006)], where they demonstrated that for any cyclic operad \(\mathcal{P}\)\ there is a certain sheaf \(\mathcal{F}\)\ asociated with it on the moduli space of stable graphs such that the Verdier dual \(D\mathcal{F}\)\ of \(\mathcal{F}\)\ is associated with the dg dual \(D\mathcal{P}\)\ of this operad, which was motivated by the relations of graph homology introduced by \textit{M. Kontsevich} [in: The Gelfand Seminars, 1990-1992. Basel: Birkhäuser. 173--187 (1993; Zbl 0821.58018)] for different kinds of Koszul dual operads, giving an important conceptual explanation of the appearance of graph cohomology of both the commutative and Lie types in computations of the cohomology of the outer automorphism group of a free group [\textit{M. Culler} and \textit{K. Vogtmann}, Invent. Math. 84, 91--119 (1986; Zbl 0589.20022); \textit{R. C. Penner}, Adv. Ser. Math. Phys. 1, 313--340 (1987; Zbl 0669.32013); \textit{A. Lazarev} and \textit{A. A. Voronov}, Adv. Math. 218, No. 6, 1878--1894 (2008; Zbl 1158.55006)]. \textit{V. Ginzburg} and \textit{M. Kapranov}'s work [Duke Math. J. 76, No. 1, 203--272 (1994; Zbl 0855.18006); Duke Math. J. 80, No. 1, 293 (1995; Zbl 0855.18007)], in which they showed that dg-duality for operads has a geometric interpretation in terms of Verdier duality for sheaves on the space of metric trees, is a precurser of Lazarev and Voronov's. The principal objective in this paper is to show that a general correspondence holds for twisted modular operads with Feynman transform and Verdier duality on the moduli space of stable metric graphs. In the course of the proof, the author establishes a relation between cyclic operads and modular operads originally proposed in the pioneering work of \textit{E. Getzler} and \textit{M. M. Kapranov} [in: Geometry, topology and physics for Raoul Bott. Lectures of a conference in honor of Raoul Bott's 70th birthday, Harvard University, Cambridge, MA, USA 1993. Cambridge, MA: International Press. 167--201 (1995; Zbl 0883.18013)]. The author's proof is almost a direct extension of the proof for cyclic operads, differing from the one in [\textit{A. Lazarev} and \textit{A. A. Voronov}, Adv. Math. 218, No. 6, 1878--1894 (2008; Zbl 1158.55006)] in three major aspects. \begin{itemize} \item[1.] The sheaves corresponding to graphs of genus greater than \(0\) are set to \(0\) in case of cyclic operads, while they are by no means \(0\) and need to be calculated as a whole in case of modular operads. \item[2.] There is no twist in case of cyclic operads, while twisted modular operads have to be dealt with. \item[3.] The author establishes a uniqueness for the correspondence, while it is not available in Lazarev and Voronov's work. \end{itemize} A simple and intuitive proof that the Feynman transform is a homotopy functor and is of a homotopy inverse is presented as an application of the correspondence.