Tilting bundles on toric Fano fourfolds (Q342851)

Consider a smooth variety \(X\) over \(\mathbb C\). A tilting object is an object \(\mathcal T\in\mathcal D^b(X)\) inducing an equivalence of bounded derived categories \(\text{RHom}_X(\mathcal T,-):\mathcal D^b(X)\simeq\mathcal D^b(A):(-){\mathop\otimes\limits^L}_A T,\) where \(A=\text{End}(\mathcal T)\). If a tilting object \(\mathcal T\) is actually a sheaf on \(X\), it is called a tilting sheaf, (or a tilting bundle if it is locally free). The main study of this article is the search for varieties where tilting objects exists, and then the possible constructions of tilting objects. For a projective curve one can construct tilting sheaves by \textit{full strong exceptional collections of sheaves} \(\{E_i\}_{i\in I}\). Given such a collection, \(\mathcal T=\bigoplus_{i\in I}E_i\) is a tilting sheaf. Conversely, if \(\mathcal T\) is a tilting sheaf, it comes from a full strong exceptional collection. The exposition is modelled on Beilinson's classical result that \(\mathcal O\oplus\mathcal O(1)\oplus\cdots\oplus\mathcal O(n)\) is a tilting bundle for \(\mathbb P^n\). Next to projective varieties are the toric varieties. On these varieties it is possible to control the collections of sheaves and to control when they are strong exceptional. Also, if they are, the endomorphism ring can be explicitly computed. This article concentrates on the particular case of toric Fano varieties. It is a finite number of such varieties in each dimension, and they are classified up to dimension \(5\). There exists a general classification algorithm. Some results have already appeared in this direction. A. King [``Tilting bundles on some rational surfaces'', \url{http://www.maths.bath.ac.uk/~masadk/papers/tilt.pdf}] has given full strong exceptional collections of line bundles for the 5 smooth toric Fano surfaces, and \textit{H. Uehara} [Int. J. Math. 25, No. 7, Article ID 1450072, 32 p. (2014; Zbl 1310.14025)] has given full strong exceptional collections of line bundles for the 18 smooth toric Fano threefolds. We state the two main result of the article verbatim: {Theorem 7.4.} Let \(X\) be one of the 124 smooth toric Fano fourfolds. Then one can construct explicitly a full strong exceptional collection of line bundles on \(X\), a database of which is contained in the computer package QuiversToricVarieties for Macaulay2. {Theorem 7.8} Let \(X\) be an \(n\)-dimensional smooth toric Fano variety for \(n\leq 4\), \(\mathcal L=\{ L_0,\dots,L_r\}\) be the full strong exceptional collection on \(X\) from the database and \(\pi:Y:=\text{tot}(\omega_X)\rightarrow X\) be the bundle map. Then \(Y\) has a tilting bundle that decomposes as a sum of the line bundles, given by \(\bigoplus_{i=1}^r\pi^\ast(L_i).\) To prove that a given collection on a toric Fano variety \(X\) is strong exceptional the author uses the construction of the not necessarily non-vanishing cohomology cones (nnnvc-cones) in the Picard lattice for \(X\) given by Eisenbud, Mustata, and Stillman [\textit{D. Eisenbud} et al., J. Symb. Comput. 29, No. 4--5, 583--600 (2000; Zbl 1044.14028)]. Then the strong exceptional condition can be verified by an algorithm, implemented in QuiversToricVarieties. It is a more involved theory to prove that a given strong exceptional collection \(\mathcal L\) on \(X\) generates \(\mathcal D^b(X).\) The author uses two methods for proving that \(\mathcal L\) is full. The first is similar to Uehara's method on the toric Fano threefolds. This uses Frobenious pushforward and obtains a set of line bundles that generates \(\mathcal D^b(X)\), and using exacts sequences with these bundles and \(\mathcal L \), \(\mathcal L\) is proved to generate \(\mathcal D^b(X)\). The second method uses the same set of line bundles to obtain a resolution of the structure sheaf of the diagonal \(\mathcal O_\Delta\). This last method includes the toric cell complex introduced by \textit{A. Craw} and \textit{A. Quintero Vélez} [Adv. Math. 229, No. 3, 1516--1554 (2012; Zbl 1280.16023)]. This is needed to obtain a minimal projective \(A-A\) bimodule resolution of the endomorphism algebra \(A=\text{End}(\bigoplus_i L_i^{-1})\). The pullback of \(\mathcal L\) to \(Y=\text{tot}(\omega_X)\) is a CY5 algebra for which the 0th, 1st, and 2nd terms of its minimal bimodule resolution is known. The symmetric structure of the CY5 algebra works to guess the 3rd, 4rt and 5th terms, the result is sheafified and restricted to \(X\) and the resulting sequence of sheaves \(S^\bullet\) is proved to be a resoulution of \(\mathcal O_\Delta\) by quiver moduli. Let \(X\) be a smooth toric Fano threefold or one of the 88 smooth toric Fano fourfoulds such that the given full strong exceptional collection \(\mathcal L\) in the database has a corresponding exact sequence of sheaves \(S^\bullet\in\mathcal D^b(X\times X)\). Let \(B\) denote the rolled up helix algebra of \(A=\text{End}(\bigoplus_{L\in\mathcal L}L^{-1})\). The calculations above leads the author to conjecture: The toric cell complex of \(B\) exists and is supported on a real four or five-dimensional torus respectively. Moreover: The cellular resolution exists in the sense of Craw-Quintero Vélez [Zbl 1280.16023], thereby producing the minimal projective bimodule resolution of \(B\), and when \(\mathcal T=\bigoplus_{L\in\mathcal L}L^{-1}\) the object \(S^\bullet\) is quasi-isomorphic to the exterior product \(T^{\vee}{\mathop\boxtimes\limits^{\mathbf{L}}}_A\mathcal T\in\mathcal D^b(X\times X).\) The birational geometry of the toric Fano fourfolds together with collections \(\mathcal L\) from a special set of line bundles (as Uehara did for threefolds) gives that the pushforward of \(\mathcal L\) onto a torus-invariant divisorial contraction is automatically full. This is a very useful technical result stated as proposition 7.2 in the article. Using this, full exceptional collections on many of the toric Fano fourfolds are obtained from the pushforward of collections on the birationally maximal examples. A database of the full strong exceptional collections on \(n\)-dimensional smooth toric Fano varieties for \(1\leq n\leq 4,\) and many of the algorithms used is contained in the QuiversToricVarieties package in Macaulay2. The article is a nice application of the theory of quivers and tilting bundles applied to toric varieties. The general theory described in various sources for projective varieties is pushed to fit in the toric setting, and is computed in the toric Fano examples. The article illustrates techniques that can be used in other (similar) settings, and the algorithms implemented in Macaulay2 can be used. Also, the article contains several important computational results on full strong exceptional collections of sheaves.