Injectives obstruct Fourier-Mukai functors
Felix Küng
TL;DR
The work develops a novel obstruction framework for lifting exact functors between derived categories to the dg/\text{A}_\infty level by replacing tilting objects with the subcategory of injectives. By computing the Inj-equivariant characteristic morphism and tracking obstructions through a divisor-based pushforward, the authors construct explicit non-Fourier-Mukai functors $\psi_\eta$, and show that for smooth hypersurfaces there exists a large space of such functors. The results yield non-liftability without requiring tilting objects and provide concrete, computable obstructions, including explicit twisted-Hodge-diamond data for divisors. The methodology potentially extends non-liftability phenomena to coherent sheaves and clarifies the landscape of functors beyond the Fourier-Mukai paradigm in geometric settings. Overall, the paper broadens the toolkit for distinguishing FM and non-FM functors via Hochschild-obstruction theory in a dg/$\text{A}_\infty$ context.
Abstract
We use injectives as a big tilting object to obstruct liftability of exact functors to the $\dg$-level. We use the inclusion of injectives into the canonical heart as a replacement for tilting objects in computations of the characteristic morphism. Then we apply this construction to proofs of non-liftability of candidate non-Fourier-Mukai functors, i.e.\ functors that do not admit an $\Ainfty$/$\dg$-lift. This approach allows explicit computation of the obstruction against an $\Ainfty$-lift. We in particular observe that this computation gives for smooth degree $d>2$ hypersurfaces an abundance of non-Fourier-Mukai functors.
