The LMO Spectrum: Factorization Homology and the E_3-Structure of the Jacobi Diagram Algebra
Takahito Kuriya
TL;DR
This work defines the LMO spectrum, a categorification of the LMO invariant via factorization homology, grounded by proving that the Jacobi-diagram algebra $\mathcal{A}_{\text{Jac}}$ carries a homotopy $E_3$-algebra structure. It derives a universal surgery formula from the excision axiom, enabling computation of the spectrum for manifolds presented by Dehn surgery, and demonstrates increased discriminatory power by constructing an $H_1$-decorated invariant that distinguishes $L(156,5)$ and $L(156,29)$, which the classical LMO invariant cannot separate. The construction links graph-complex formality, the Drinfeld associator, and the Kashiwara–Vergne problem to provide a coherent algebraic backbone for factorization-homology-based 3-manifold invariants, including an operator-state formalism that yields a universal $H_1$-decorated weight system. The results open paths to higher-order invariants via Massey products and Johnson homomorphisms and suggest parameterized (character-variety–dependent) refinements that could address LMO completeness. Overall, the paper provides a rigorous, computable, and potentially more powerful framework for 3-manifold invariants rooted in higher-algebraic structures.
Abstract
We define the LMO spectrum, a categorification of the Le-Murakami-Ohtsuki (LMO) invariant for 3-manifolds, using factorization homology. The theoretical foundation is our main algebraic result (Theorem A): the algebra of Jacobi diagrams, $\AJac$, possesses a homotopy $E_3$-algebra structure. This is a necessary condition for consistency within factorization homology, and the proof relies on the formality of the little 3-disks operad. A universal surgery formula is derived from the excision axiom (Theorem B), providing a computational basis independent of conjectural models. As an application (Theorem C), we construct an ``$H_1$-decorated LMO invariant'' that distinguishes the lens spaces $L(156, 5)$ and $L(156, 29)$, a pair that the classical LMO invariant fails to separate.
