The Johnson homomorphism, embedding calculus and graph complexes
Florian Naef, Thomas Willwacher
TL;DR
The paper connects the Johnson homomorphism and its cokernel to graph complexes arising from Goodwillie-Weiss embedding calculus, providing a cohesive framework that unifies these invariants with higher-loop Enomoto-Satoh traces. It proves stable injectivity of the Johnson homomorphism in weight regimes $g\ge 3W\ge 3$, and gives an explicit, representation-theoretic decomposition of the cokernel in terms of $\mathrm{Sp}(2g)$-representations with multiplicities $M_{\mu}^W$, computed via top-weight Euler characteristics $\chi_{h,\lambda}$ of moduli spaces and branching data. The approach leverages a loop-order filtration and a spectral sequence to identify higher ES traces whose images detect the cokernel, and it provides concrete calculations up to weight $6$, showing where higher-loop contributions first appear. Finally, the work situates these graph complexes within the embedding calculus framework, offering a conceptual link to diffeomorphism groups and the rational Torelli-like structure of surface automorphisms.
Abstract
We explain how the Johnson homomorphism and the Enomoto-Satoh trace, as well as higher-loop-order generalizations, can be obtained from graph complexes originating in the Goodwillie-Weiss calculus. This paper can be seen as an addendum to our earlier work. It contains little new mathematical content, but is intended to give an overview of a different viewpoint on the Johnson homomorphism, for experts working mainly in the latter area.
