Knot invariants and new weight systems from general 3D TFTs (Q2428147): Difference between revisions

Following Kontsevich, the perturbative calculation of the Chern-Simons path integral can be seen as a pairing between chains and cochains. Both are motivated by Feynman diagrams with cochains coming from the topological data of the \(3\)--manifold under consideration, while the chains coming from so--called weight systems usually constructed from Lie algebra data. The fact that the chains turn out to be cycles and cochains turn out to be cocycles explains the topological invariance of the path integral, and imposes a restriction on weight systems. Bar-Natan originally conjectured that all possible weight systems come from semisimple Lie algebras, but this was later disproved by Vogel. Vogel's proof does not provide an explicit construction of alternative weight systems, but Rozansky and Witten gave (what is believed to be) one associated with holomorphic vector bundles on hyper--Kähler manifolds. There is no rigorous proof that the Rozansky--Witten systems are not Lie algebraic though. In this paper the authors give a generalization of the Rozansky--Witten construction based on representations of \(Q\)--structures (\(L^\infty\) algebras) on graded manifolds. A \(Q\)--structure is a degree \(1\) vector field with \(Q^2=0\), and its representation is an extension of \(Q\) to a vector bundle over the same manifold. The resulting knot invariants are \(H_Q\) valued, in the case of Rozansky-Witten \(Q=\overline{\partial}\), the Dolbeault operator. The construction is based on the Alexandrov--Kontsevich--Schwarz--Zaboronsky perturbative framework for topological field theories complemented by Batalin-Vilkovisky formalism for Wilson loops. In this framework the cycle condition, which comes from the Jacobi identity in the Lie algebraic setting, is derived from a more general relation for \(Q\)--structures. The authors define an extended Chevalley-Eilenberg chain complex of Hamiltonian vector fields, and show that it is isomorphic to the usual extended graph complex built from Feynman diagrams. This generalizes a result of Kontsevich to the case with Wilson loops. The differential of the Chevalley-Eilenberg complex comes from a single Ward identity in the Batalin-Vilkovisky formalism. Working with this complex the authors verify heuristically that \(Q\)--structures do indeed yield knot invariants. An example based on holomorphic vector bundles on hyper--Kähler manifolds is worked out explicitly, the main difference with Rozansky--Witten being that the invariants are \(H_{\overline{\partial}}\) valued rather than complex valued. This allows the authors to remove some restrictions on the type of hyper--Kähler manifolds used to produce a weight system. It remains open whether these new weight systems are not Lie algebraic, or even if the topological field theories involved can be fully quantized non-perturbatively, i.e. if they lead to some knot polynomials. As a first step the authors suggest obtaining a skein relation.