Orlov spectra: bounds and gaps (Q1762414)

Bounded derived categories of complexes of coherent sheaves on algebraic varieties have to be considered as a very deep algebraic and categorical tool to study geometrical properties, in particular the birational behavior of (smooth) projective varieties. Such categories always carry a triangulated structure. In this paper, the authors consider a new invariant of a triangulated category, the \textit{Orlov spectrum}, defined by \textit{D. Orlov} [Mosc. Math. J. 9, No. 1, 143--149 (2009; Zbl 1197.18004)]. Given a triangulated category \(\mathcal T\) and a full subcategory \(\mathcal I\), one can consider \(\langle {\mathcal I} \rangle_0\) to be the full subcategory of all the direct summands of finite coproducts of shifts of objects in \(\mathcal I\). Then one can define inductively \(\langle {\mathcal I} \rangle_n\) to be the category of all the objects \(B\) fitting a triangle \(B_{n-1} \to B \to B_0\), where \(B_i\) belongs to \(\langle {\mathcal I} \rangle_i\). An object \(E\) is a \textit{generator} of \(\mathcal T\) if there exists an integer \(n\) such that \(\langle E \rangle_n={\mathcal T}\). The \textit{generation time} of a generator \(E\) is the minimal integer with such a property. The \textit{Orlov spectrum} of \(\mathcal T\) is the subset of \(\mathbb{N}\) given by generation times of all possible generators of \(\mathcal T\). The infimum (resp. the supremum) of the Orlov spectrum is called \textit{Rouquier dimension} (resp. \textit{ultimate dimension}) of \(\mathcal T\). The \textit{gaps} in the Orlov spectrum are the missing integers. In this paper, the authors state and consider conjectures relating Orlov spectra, bounds on ultimate and Rouquier dimension, and gaps to geometrical problems. Roughly: 1) For a smooth algebraic variety \(X\) the Rouquier and the Krull dimension are expected to be equal, 2,3) The length of any gap in the Orlov spectrum of \(D^b(X)\) is controlled by semiorthogonal decompositions. The main interest of these conjectures relies on their corollaries, in particular, they would imply that gaps of length \(n\) and \(n-1\) in \(D^b(X)\) are birational invariants of a smooth projective variety \(X\) of dimension \(n\). This leads to a conditional rationality criterion: if \(X\) is rational then the length of any gap in the spectrum of \(D^b(X)\) cannot exceed \(n-2\). The starting point for these conjectures is a theorem of Orlov [loc. cit.] which states that \(\{1,2\}\) is always contained in the Orlov spectrum of a curve, with equality if and only if the curve is rational. The authors give a careful introduction and description of these new concepts, and prove results providing new evidences for the very challenging conjectures stated above. In particular, they provide an upper bound for the ultimate dimension of the triangulated category of singularities associated to an isolated hypersurface singularity. From this, they can describe the whole spectrum in the case of an \(A_n\) singularity. Moreover, suppose \(G\) is a generator of a triangulated category \(T\), and there is a semiorthogonal decomposition such that \(G\) can be split as a direct sum of objects \(G_i\) in each component. The length of the gaps in the subset of the spectrum of \(T\) corresponding to such generators is shown to be bounded by the generation times (for the corresponding component) of the \(G_i\)'s. An explicit calculation for the upper bound of the generation time of a full exceptional collection is also provided. Moreover, the authors consider also Calabi-Yau projective hypersurfaces, and provide generators and their generation times using Seidel-Thomas twists. They can also prove, using spherical objects and braid group actions, that the ultimate dimension of the derived Fukaya category of a symplectic surface of genus \(g\) is at least \(4g\), and that the Orlov spectrum of \(D^b(C)\) is \(\{1,2,3,4\}\) if \(C\) is a smooth elliptic curve.