On the extended Whittaker category (Q1740465)

This paper is a part of the literature on the geometric Langlands conjecture that seeks to prove the equivalence \(\text{IndCoh}_{\mathcal{N}}(\text{LocSys}_{\check{G}})\simeq \mathfrak{D}(\text{Bun}_G)\) (proposed in this form in [\textit{D. Arinkin} and \textit{D. Gaitsgory}, Sel. Math., New Ser. 21, No. 1, 1--199 (2015; Zbl 1423.14085)]) by embedding the DG-categories on either side of the equivalence into a larger ambient DG-category and matching a collection of generators for each category under these embeddings. The primary contributions of this paper to this program are as follows: (1) A construction of this larger ambient category as the \textit{extended Whittaker category} \(\mathcal{W}h(G,\text{ext})\), (2) a construction of the \textit{functor of Whittaker coefficients} \(\text{coeff}_{G,\text{ext}}:\mathfrak{D}(\text{Bun}_G)\to\mathcal{W}h(G,\text{ext})\), and (3) a proof for \(G=\mathrm{GL}_n\) and \(G=\mathrm{PGL}_n\) of the conjecture of Gaitsgory that the functor of Whittaker coefficients is fully faithful (see [\textit{D. Gaitsgory}, Astérisque 370, 1--112 (2015; Zbl 1406.14008)] for this conjecture together with the proposed strategy of proof). The Ran space of the curve \(X\) -- the moduli of finite non-empty subsets of \(X\) -- plays a central role in the constructions and proofs of this paper, many of which apply more generally than to the specific examples involved in the geometric Langlands conjecture. For instance, for any \(\mathfrak{D}(\text{Ran})\)-module category \(\mathcal{C}\) equipped with the action of a certain twisted group, the author constructs a Ran version of the extended Whittaker category for \(\mathcal{C}\) over \(\text{Ran}\). Additionally, the author constructs an `independent subcategory' functor that assigns to certain modules over \(\mathfrak{D}(\text{Ran})\) (called `naive unital') an underlying independent DG-category. \(\mathcal{W}h(G,\text{ext})\) is then built by taking the independent subcategory of the Ran version of the extended Whittaker category for \(\mathfrak{D}\)-modules on the \(\Omega_X\)-twisted Grassmannian of \(X\). Similarly, the construction of the functor of Whittaker coefficients proceeds by first defining such a functor over the Ran space via a procedure of `averaging over an action of meromorphic jets of functions to a unipotent group, and then extracting an independent version of this functor. The proof that the coefficient functor is fully faithful for \(G=\mathrm{GL}_n\) or \(\mathrm{PGL}_n\) is presented in the final section and relies on what the author considers ``the main new idea of the present paper'', namely a ``blow-up trick'' that allows for a clever induction on \(n\). Finally, I would be remiss not to mention that the introduction of this paper contains a marvellous discussion of the analogy between the contents of this paper and the function theoretic version of the Langlands program. The function theoretic version involves the analysis of vector space of automorphic functions via an attempted decomposition into Whittaker coefficients, and the appearance of certain geometric objects is demystified by the function theoretic analogy (e.g.\ the appearance of adèles and adèlic group points in the function theoretic program leads to the Ran space and jets of group valued meromorphic functions in the geometric program).