2-hereditary algebras and almost Fano weighted surfaces (Q515590)

When an algebra is the endomorphism ring of a tilting bundle \(\mathcal T\) on a sufficiently nice weighted projective variety \(\mathbb X\) there is an equivalence of derived categories \(D^b_{fg}(\text{End}(\mathcal T))\simeq D^b_c(\mathbb X)\). This says that the representation theory of \(\text{End}(\mathcal T)\) can be studied, upto derived equivalence, by the representation theory of the projective space \(\mathbb X\). The main examples are Ringel's canonical algebras \(\Lambda\). These are \(\text{End}(\mathcal T)\) for a tilting bundle \(\mathcal T\) on a weighted projective line \(\mathbb X\). If \(\mathbb X\) is Fano, and only then, there exists a tilting bundle \(\mathcal T^\prime\) such that \(\text{End}(\mathcal T)\) is hereditary. Thus \(\Lambda\) is derived equivalent to an hereditary algebra if and only if \(\mathbb X\) is Fano. This article considers the generalization of this to the case of weighted projective surfaces. This implies the generalization of hereditary given by Herschend-Iyama-Opperman to the notion of \(n\)-hereditary, and it implies the different ways of studying weighted projective spaces and varieties. These are considered as log varieties of the form \((X,\Delta=\sum_i(1-\frac{1}{p_i})D_i)\) or as the associated \textit{Geigle-Lenzing orders} defined by Iyama and Lerner. The log variety viewpoint allows to consider geometric concepts such as (almost) Fano and the log minimal program, and the order approach gives a category of coherent sheaves allowing the definition of tilting bundles for weighted projective spaces and giving Serre duality. Now the main question under investigation is which projective weighted surfaces \(\mathbb X\) that have tilting bundles \(\mathcal T\) such that \(\text{End}(\mathcal T)\) is \(2\)-hereditary. These bundles are named \textit{2-hereditary}. Hille and Perling have proved that all rational surfaces have tilting bundles, and it is an open question if these are the only ones. The study in this article conjectures that a weighted projective surface \(\mathbb X\) has a 2-hereditary tilting bundle if and only if the weighted divisors \(D_i\) are rational and \(\mathbb X\) is almost Fano. This conjecture is related to a conjecture of Herschend, Iyama, Minamoto, and Oppermann implying that the Fano projective spaces of dimension \(n\), weighted on hyperplanes, are the ones which have an \(n\)-hereditary tilting bundle. The author constructs a 2-hereditary tilting bundle on the non-Fano Hirzebruch surface \(\mathbb F_2\), proving that \textit{almost Fano} cannot be dropped from the conjecture. To support it, the article contains the following main results, stated verbatim: Theorem 1.2 Let \((X,\Delta)\) be a weighted projective surface with a 2-hereditary tilting bundle. Then \(-(K_X+\Delta)\) is nef and \((K_X+\Delta)^2\geq 0\). In particular, if \(\Delta=0\), then either \(X\) is almost Fano, or it is a blowup of \(\mathbb P^2\) at 9 points in almost general position. Theorem 1.3 Let \(X\) be a Fano surface of degree \(K_X^2\geq 3\). Then \(X\) has a 2-hereditary tilting bundle (explicitly described) which is a direct sum of line bundles. The author gives the necessary definitions, and explain the various standard ways to look for a tilting bundle \(\mathcal T\). This standard ways are generalized to the purpose of finding \(n\)-hereditary tiling bundles. The easiest way is to look for direct sums of line bundles. The generating condition can then be checked as Lerner-Oppermann give a set of generators for the derived category of a weighted projective variety. The other conditions are checked by cohomology computations which are not too hard for line bundles. In the dimension one case, the tilting bundles can be obtained by either mutating other tilting bundles or using group actions. These methods are then generalized to find examples of 2-hereditary tilting bundles and their associated 2-hereditary algebras. Computing examples, the author finds that the following Fano weighted projective surfaces have 2-hereditary tilting bundles: \(\mathbb P^1\times\mathbb P^1\) weighted on a smooth \((1,1)\)-divisor, \(\mathbb P^2\) weighted on a conic with weight \(2\), and \(\mathbb P^2\) weighted on \(4\) lines in general position with all weights equal to \(2\). The computations in this article starts with recalling the approach given by Iyama and Lerner of studying weighted projective varieties via Geigle-Lenzing (GL)-orders. This gives a simple necessary condition for a weighted projective variety to have an \(n\)-hereditary tiling bundle. In general, the introduction of orders is a way to associate a category of coherent sheaves to a weighted projective variety, and this is very nicely illustrated in the article. It gives the interplay between representation theory on one side, and the projective geometry on the other side. The standard definitions and methods are recalled. The \textit{Cox ring} is useful as a tool for for calculation of endomorphism algebras of tilting bundles and \(\text{Ext}\)-groups. The author gives a version of the Cox ring for the GL-order of interest. In defining the \textit{almost Fano weighted projective varieties} some basics of the \textit{log minimal model program} are recalled in a very suitable way. Some \(n\)-hereditary tilting bundles appear as associated to group quotients, and the relation to the weighted spaces are given. With this, the author can give the generators of the derived category and give examples of \(2\)-hereditary tilting bundles on weighted projective surfaces. This is a very nice article, explaining the interplay between algebra (representation theory) and projective geometry. The article is very well written, nearly self-contained, and it contains a lot of results explaining when this interplay can be used.