Equivariant-constructible Koszul duality for dual toric varieties (Q2369020)

Start with an affine toric variety \(X\) (defined by a cone \(\sigma\) and its faces) and its dual \(X^\vee\) (defined by \(\sigma^\vee\)). At the crudest level, there is hope to relate equivariant sheaves on \(X\) with complexes of constructible sheaves on \(X^\vee\). One might, for instance, ask for an exact equivalence of bounded derived categories. In reality the situation is considerably more complicated. What is constructed here is an equivalence \(K\) of triangulated categories called \(D^b({\mathcal A}-\bmod_f)\) and \(D^b(LC_{\mathcal F}(X^\vee))\), which is not \(t\)-exact but is a Koszul equivalence. Very approximately this means that simple objects on the left map to injectives and simple objects on the right come from projectives. As for the categories, they are mixed versions of slightly more elementary categories, which are recovered by forgetting a grading. The ungraded version of \(D^b({\mathcal A}-\bmod_f)\) is just \(D^b_T(X)\), the \(T\)-equivariant bounded derived category of sheaves on \(X\). The ungraded version of \(D^b(LC_{\mathcal F}(X^\vee))\) is the derived category of a category of locally constant unipotent sheaves on \(X^\vee\), meaning that on any \(T\)-orbit \(O\) the restriction is a local system and the monodromy action of \(\pi_1(O)\) is unipotent on any stalk. The style of the paper is of necessity highly technical and even to explain the notation would exceed what can well be done in a review. The authors are aware of the difficulty: they relegate (their word) some technical proofs to a separate section, and provide a lengthy but still sketchy introduction. Despite this, the paper is really readable only by experts.