Twisted rings and moduli stacks of ``fat'' point modules in non-commutative projective geometry (Q765647)

In noncommutative projective geometry, one studies a noncommutative connected graded algebra \(A\). The simple objects in the category \(A\)-gr of graded \(A\)-modules does not give sufficient information about \(A\), and one has to pass to the category of \(A\)-tails \(\text{Proj}(A):=A-\text{Gr}/\text{tors}\) where tors is the subcategory of \(A_{>0}\)-torsion modules. In this article the authors introduce Artin stacks to study the moduli of more general simple objects called dubbed ``fat'' points. Starting with a noetherian graded \(k\)-algebra, finitely generated in degree one which satisfies good homological properties, assuring the existence of Hilbert schemes à la Artin and Zhang. For an integer \(m\geq 1\), an \(m\)-point is an \(A\)-tail \(M\) such that the corresponding saturated \(A\)-module has constant Hilbert function \(m\), i.e. \(\dim_k H^0(M(i))=m\) for all \(i\in\mathbb Z\). The point modules are then essentially the \(1\)-points, while the \(m>1\) case correspond to what Artin has called a ``fat'' point. The authors prove that \(m\)-points are parametrised by an Artin stack. The simple \(m\)-points form an open substack \(\mathcal Y\) constructed by studying an appropriate Hilbert scheme. There is an open subset \(H\) of projective Hilbert scheme with free action of \(PGL_m\) on it such that \(\mathcal Y\) is isomorphic to the quotient stack \([H/GL_m]\) where the \(PGL_M\)-action is trivially lifted to \(GL_m\). For point modules, the stack is \([H/\mathbb G_m]\), \(\mathbb G_m\) acting trivially. By considering \(m\)-points, the article generalizes methods from Artin, Tate, and Van den Bergh. Let \(Y\) be the Hilbert scheme of point modules, and let \(\mathcal M\) be the universal point module on \(Y\). The universal property ensures that the shift functor induces an automorphism \(\sigma\) of \(Y\), such that \(\mathcal M(1)\simeq\sigma^\ast\mathcal M\otimes_Y\mathcal L\) for some line bundle \(\mathcal L\) on \(Y\). The twisted sheaf of rings \(\mathcal E:=\mathcal O_Y\oplus\mathcal L\oplus(\sigma^\ast\otimes_Y\mathcal L)\oplus\cdots,\) is useful for studying \(A\) as there is a canonical map from \(A\rightarrow H^0(\mathcal E)\). \(\text{Proj} H^0(\mathcal E)\) is essentially the maximal commutative subscheme of \(\text{Proj}(A)\). The author seek a similar construction for \(m\)-points. There is a universal \(m\)-point \(\mathcal M\) on \(\mathcal Y\) and an automorphism \(\sigma:\mathcal Y\rightarrow\mathcal Y\) induced by the shift functor such that \(\mathcal M(1)\simeq\sigma^\ast\mathcal M\). To this data it is associated a twisted ring by passing to an Azumaya algebra. The coarse moduli space of \(\mathcal Y\) is the separated algebraic space \(Y=H/PGL_m.\) The \(PGL_m\)-torsor \(p:H\rightarrow Y\) corresponds to an Azumaya algebra \(\mathcal A\) on \(\mathcal Y\) which is the \(m\)-point analogue of \(\mathcal O_Y\). The automorphism \(\sigma\) corresponds to an invertible \(\mathcal A\)-bimodule \(\mathcal B\) which takes the role of \(\mathcal L_\sigma\) in the point scheme case. It is thus possible to form the twisted sheaf of rings \(\mathcal E\) as before, and there is a canonical morphism \(A\rightarrow H^0(\mathcal E)\). The twisted ring \(H^0(\mathcal E)\) is thought of as capturing the stratum \(\text{Proj}A\) which is PI of PI-degree exactly \(m\). In particular, if \(\text{Proj}A\) arises from an order on a variety, then one expects that for appropriate \(m\), \(Y\) is the Azumaya locus of the order and \(\mathcal A\) is the restriction of that order to \(Y\). Rogalski and Zhang show that the canonical morphism is essentially surjective, which do not hold for \(m>1\). Part of the problem is that only simple \(m\)-points are considered, and so the information about \(m^\prime\)-points for \(m^\prime<m\) is lost. Considering all \(m\)-points, not only the simple ones, gets around this problem, but it is not known how to associate an order to such a moduli stack. The key in this construction is the fact the moduli stack of simple \(m\)-points constructed here has a very special form \([H/GL_m]\) to which is naturally associated an Azumaya algebra. The interest in studying \(m\)-points stems from Artins's conjecture on the classification of non-commutative surfaces. The article's main result states that a non-commutative surface is birationally PI if it has a surface worth of simple \(m\)-points. This reinforces the well-observed phenomenon that points disappear the more non-commutative one gets. Finally, the author gives a local criterion for when a non-commutative projective surface has a surface worth of simple \(m\)-points. The last sentence in the introduction says that this last question could be approached by deformation theory, and this is a good idea. Some expository material is included: A relationship between Azumaya algebras and certain stacks is recalled. The stacks are actually constructed, in particular the stack of simple \(m\)-points.This part includes the definition of (this version of) non-commutative projective geometry, with the necessary conditions imposed on the algebra \(A\). The article gives a nice generalization of the point-modules to fat point modules, and the application to the birational classification of non-commutative surfaces is of great value.