In http://arxiv.org/abs/math/0405089 Seidel and Smith constructed a link invariant using Lagrangian Floer theory that was conjectured to be equivalent to Khovanov homology. The equivalence was recently proved by Abouzaid and Smith in http://arxiv.org/abs/1504.01230. Their construction involves associating a Hamiltonian diffeomorphism $\phi_B$ with a braiding between two crossingless matchings C and C', which are in turn associated with Lagrangian submanifolds $L_C,L_{C'}$ of a particular symplectic manifold. Then the invariant of the resulting link is defined to be $HF^*(L_C,\phi_B(L_{C'}))$, the Floer cohomology of the pair $L_C,L_{C'}$. Modulo a few additional details this is the extent to which I understand the theory. Given that there are a number of starting points for understanding the Jones polynomial (e.g. state sums, quantum group intertwiners etc.), is there a natural symplectic geometry-based explanation that may first help me to understand a "decategorified" version of symplectic Khovanov homology? I have no idea why all the different ingredients make a sensible starting point for defining something that eventually turns out to give the correct answer.


One thing you didn't mention is the manifold in which this calculation happens: the Slodowy slice to a nilpotent of type $(n,n)$. This manifold is the key to everything, since it is a geometric avatar of the invariants inside the representation $(\mathbb{C}^{2})^{\otimes 2n}$.

In what sense is this true? First, we have an embedding of this tensor product into $\bigwedge{}^{\! 2n}(\mathbb{C}^{2}\otimes \mathbb C^{2n})$ as the elements of weight 0 under $\mathfrak{sl}(2n)$. By skew Howe duality, the invariants are the intersection of this weight space with the unique copy of the simple representation of $\mathfrak{sl}(2n)$ with highest weight $(2,\dots, 2, 0, \dots, 0)$. The Slodowy slice under discussion is actually the Nakajima quiver variety for this weight space, so that's one level of geometric avatarage; the Hamiltonian diffeomorphisms correspond to the action on invariants of the braiding for quantum groups, and the Lagrangians for crossingless matchings to the invariant vector for that matching.

This actually fits into a much more general picture: Lauda, Queffelec and Rose have proven that you always get Khovanov homology out of a category when you have a categorical action of $\mathfrak{sl}(2n)$ categorifying the simple with highest weight $(2,\dots, 2, 0, \dots, 0)$. Nakajima's construction of the action of $\mathfrak{sl}(2n)$ on the cohomology of the quiver varieties for this representation has an obvious candidate for a lift to the Fukaya category: he pushes and pulls on some Lagrangian correspondences, now just think of them as Lagrangian correspondences and look at the induced functor on the Fukaya category. There's a deformation quantization version of this that really works (http://front.math.ucdavis.edu/1208.5957), but the Fukaya category is a much spookier place. Seidel-Smith is in all likelihood just implementing the LQR construction there; of course, good luck to whoever wants to check that those correspondences really give a categorical action.

| cite | improve this answer | |

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.