Ryan Cohen Reich | Searcharxiv

arXiv2014

Obvious natural morphisms of sheaves are unique

We prove that a large class of natural transformations (consisting roughly of those constructed via composition from the "functorial" or "base change" transformations) between two functors of the form $\cdots f^* g_* \cdots$ actually has only one element, and thus that any diagram of such maps necessarily commutes. We identify the precise axioms defining what we call a "geofibered category" that ensure that such a coherence theorem exists. Our results apply to all the usual sheaf-theoretic contexts of algebraic geometry. The analogous result that would include any other of the six functors remains unknown.

math.AG↗

arXiv2013

Twisted geometric Satake equivalence via gerbes on the factorizable grassmannian

We prove a generalization of the twisted geometric Satake equivalence of Finkelberg--Lysenko in the context of the factorizable grassmannian of a reductive group G relative to a smooth curve X, similar to Gaitsgory's generalization in "On de Jong's conjecture" of the Satake equivalence itself. In order to express the dependence of this result on factorizability, we consider a certain 2-category of gerbes on this grassmannian, which we call "sf gerbes"; the theorem then concerns perverse sheaves twisted by such gerbes. This notion of twisting differs from, but in special cases is equivalent to, that considered by Finkelberg--Lysenko. We give a classification of these gerbes in terms of quadratic forms on the coweight lattice of G and gerbes on X. In addition, we introduce the technical device of universal local acyclicity (ULA) into the proof of the Satake equivalence, using it in combination with Beilinson's nearby cycles gluing of perverse sheaves to reduce all of the main constructions in the Mirkovic--Vilonen proof of the equivalence to the ULA case, where they have simple descriptions whose properties have simple proofs. We also give a geometric proof that the category of spherical perverse sheaves on the ordinary affine grassmannian is semisimple.

math.RT↗