Height of exceptional collections and Hochschild cohomology of quasiphantom categories (Q890769)

One of the very interesting recent developments in the theory of derived categories of coherent sheaves on smooth projective varieties was the discovery of so-called phantom and quasi-phantom categories. If \(X\) is a smooth projective variety then an admissible subcategory \(\mathcal{A}\) of \(D(X)\), the derived bounded category of coherent sheaves on \(X\), is a phantom (resp. quasi-phantom) its Hochschild homology and cohomology are zero and its Grothendieck group is trivial (resp. torsion). At the time this paper was written there were three known examples of families of surfaces whose derived categories contain a quasi-phantom subcategory [\textit{C. Böhning} et al., Adv. Math. 243, 203--231 (2013; Zbl 1299.14015)], [\textit{V. Alexeev} and \textit{D. Orlov}, Math. Ann. 357, No. 2, 743--759 (2013; Zbl 1282.14030)], [\textit{S. Galkin} and \textit{E. Shinder}, Adv. Math. 244, 1033--1050 (2013; Zbl 1408.14068)] and one family whose derived categories contains a phantom subcategory [\textit{C. Böhning} et al., J. Eur. Math. Soc. (JEMS) 17, No. 7, 1569--1592 (2015; Zbl 1323.14014)]. In these examples, the (quasi-)phantom subcategory is the orthogonal complement to an exceptional collection of the maximum possible length which moves with the surfaces in moduli but does not deform. So a point of view one could take is that the presence of a (quasi-)phantom category accounts for the failure of a Torelli-type theorem to hold. The main application of the present paper is to show that for these examples, the deformation theory of \(D(X)\) is canonically isomorphic to the deformation theory of its (quasi-)phantom subcatory. This is made precise as follows. Let \(\mathcal{D}\) be a DG-category and \(\mathcal{A}\) a full DG-subcategory. Then there is a restriction morphism \(\mathrm{HH}^\bullet(\mathcal{D}) \to \mathrm{HH}^\bullet(\mathcal{A})\). This restriction morphism is a morphism of Gerstenhaber algebras and therefore induces a morphism of formal deformation spaces. So it is desirable in general to understand the mapping cone for the restriction morphism. If the restriction morphism is an isomorphism in degrees \(0\) through \(3\) then restriction identifies the formal deformation spaces. In the especially clear and well-written paper under review, the author studies a smooth projective variety \(X\) and the Čech DG-enhancement of \(D(X)\) equipped with a semi-orthogonal decomposition \(\mathcal{D} = D(X) = \langle \mathcal{A}, \mathcal{B} \rangle\). The author associates a complex \(\mathrm{NHH}^\bullet(\mathcal{B},\mathcal{D})\), called the normal Hochschild cohomology. The normal Hochschild cohomology turns out to fit into an exact triangle \[ \mathrm{NHH}^\bullet(\mathcal{B},\mathcal{D}) \to \mathrm{HH}^\bullet(\mathcal{D}) \to \mathrm{HH}^\bullet(\mathcal{A}) \to \] When \(\mathcal{B}\) is generated by an exceptional collection, the author shows that \(\mathrm{NHH}^\bullet(\mathcal{B},\mathcal{D})\) can be computed by a natural spectral sequence involving the ext-spaces among the terms in the exceptional collection. From this he derives a criterion for the vanishing of \(\mathrm{NHH}^\bullet(\mathcal{B},\mathcal{D})\) in low degrees, which is used to define height and pseudoheight for exceptional collections, then applied to the examples of surfaces with (quasi-)phantoms.