Knot homology via derived categories of coherent sheaves. II: \(\mathfrak{sl}_{m}\) case (Q944254)

Let \(\mathfrak g\) be a complex semisimple, finite-dimensional Lie algebra and let \(U_q(\mathfrak g)\) be the corresponding quantized enveloping algebra, \(q\) generic. Let \(\lambda\) be a dominant integral weight and let \(V_\lambda\) be the irreducible highest weight \(U_q(\mathfrak g)\)-module with highest weight \(\lambda\). If \(\underline{\lambda} = (\lambda_1, \dots, \lambda_n)\) is an \(n\)-tuple of dominant integral weights, then we set \(V_{\underline{\lambda}} = V_{\lambda_1} \otimes \dots \otimes V_{\lambda_n}\). Given a tangle \(T\) with strands labeled by dominant integral weights, say \(\underline{\lambda} = (\lambda_1, \dots, \lambda_n)\) on the top and \(\underline{\mu} = (\mu_1, \dots, \mu_n)\) on the bottom, there is a map \(\psi(T): V_{\underline{\lambda}} \to V_{\underline{\mu}}\) constructed in [\textit{N. Yu. Reshetikhin} and \textit{V. G. Turaev}, Commun. Math. Phys. 127, No. 1, 1--26 (1990; Zbl 0768.57003)]. In this way, versions of well-known link invariants are recovered. The categorification in this context, proposed in [\textit{M. Khovanov}, Duke Math. J. 101, No. 3, 359--426 (2000; Zbl 0960.57005), Algebr. Geom. Topol. 2, 665--741 (2002; Zbl 1002.57006)] consists of a) choosing a triangulated category \(D_{\underline{\lambda}}\) for every \(n\)-tuple of dominant integral weights, and b) defining a map \(\Psi\) from the set of \((\underline{\lambda}, \underline{\mu})\)-tangles to the set of isomorphism classes of exact functors \(D_{\underline{\lambda}} \to D_{\underline{\mu}}\), in a such a way that the original calculus of Reshetikhin and Turaev has to be recovered at the level of the Grothendieck groups, that is \(K(D_{\underline{\lambda}}) \simeq V_{\underline{\lambda}}\) and \(\psi(T)\) coming from \(\Psi(T)\). Thus, the question is whether such categorifications do exist, for all \(\mathfrak g\) and all \(\lambda_i\). In the case \(\mathfrak g = \mathfrak{sl}_2\) and all \(\lambda_i\) equal to the natural weight, the answer was given by Khovanov in [loc. cit]; while \(\mathfrak g = \mathfrak{sl}_n\) and all \(\lambda_i\) minuscule was done in [\textit{M. Khovanov} and \textit{L. Rozansky}, Fundam. Math. 199, No. 1, 1--91 (2008; Zbl 1145.57009)]; [\textit{J. Sussan}, Category \(\mathcal O\) and \(\mathfrak{sl}(k)\) link invariants, \url{math.QA/0701045}]. In this paper, the authors remain in the setting \(\mathfrak g = \mathfrak{sl}_n\) and all \(\lambda_i\) minuscule; they consider \(D_{\underline{\lambda}}\) as the derived category of \(\mathbb C^{\times}\)-equivariant coherent sheaves on the variety Gr\(_{\underline{\lambda}}\), that is the convolution product of certain orbits Gr\(_{\lambda_i}\) in the affine Grasmannian. To a tangle \(T\) as above they attach a functor \(\Psi(T)\) and check that its class is an invariant, as desired. The choice of \(D_{\underline{\lambda}}\) comes from the geometric Satake correspondence. Conjecturally, the Khovanov-Rozansky homology is recovered in this way. Also, a relation with \textit{C. Manolescu} [Adv. Math. 211, No. 1, 363--416 (2007; Zbl 1117.57012)] is pointed out.