An isomorphism of motivic Galois groups (Q529231)

The classical (conjectural) theory of motives, as envisioned by Grothendieck (and refined by Beilinson), predicts that there is for every field \(k\) (for the purpose of this review, as in the article under review, \(k\) will have characteristic zero) a \(\mathbb{Q}\)-linear abelian category \(MM(k)\) of ``mixed motives'', which receives a universal (co)homology theory from quasi-projective varieties over \(k\). The category \(MM(k)\) is supposed to be \textit{neutral Tannakian}; this means that it is closed symmetric monoidal, all objects are strongly dualizable, and there is a conservative functor to (perhaps graded) vector spaces over an extension of \(k\). It would then follow from Tannakian reconstruction theory that \(MM(k)\) is equivalent to to the category of representations of a (uniquely determined) pro-algebraic group \(G(k)\), known as the motivic Galois group. There are two directions from which this conjectural theory can be approached. If \(k \subset \mathbb{C}\) then Nori has constructed a \(\mathbb{Q}\)-linear abelian category \(HM(k)\) which is, in a precise sense, a universal factorization of the Betti realisation functor sending an algebraic variety to the Betti cohomology of its complex points. The category \(HM(k)\) is Tannakian. Moreover, if \(MM(k)\) exists and satisfies the expected properties, then it must be equivalent to \(HM(k)\). Consequently the construction of \(HM(k)\) yields in particular an unconditional definition of the motivic Galois group \(G(k)\). The downside is that \(HM(k)\) is, for many problems, impossibly hard to work with. The second approach, due to Voevodsky and others, attempts to construct directly the (bounded) derived category of \(MM(k)\). The construction thus yields a triangulated category \(DM^{gm}(k, \mathbb{Q})\) which, under the assumption of certain standard conjectures, carries a \(t\)-structure with heart \(MM(k)\) (and is under the same assumptions equivalent to the bounded derived category). Using a triangulated Betti realisation functor, Ayoub has constructed a pro-algebraic group \(G_A(k)\) associated with \(DM(k, \mathbb{Q})\). (He uses a kind of derived weak Tannakian reconstruction theory.) It can be shown that, if all the conjectures hold, then not only \(DM^{gm}(k, \mathbb{Q}) = D^b(HM(k))\), but also \(G_A(k) = G(k)\). In the article under review, the authors establish this latter isomorphism unconditionally.