Homogeneous spaces in tensor categories
Kevin Coulembier, Alexander Sherman
TL;DR
The paper addresses the problem of defining and understanding homogeneous spaces within tensor categories of moderate growth by proving the existence of quotients $\mathcal{G}/\mathcal{H}$ for algebraic groups $\mathcal{G}$ with a subgroup $\mathcal{H}$. It develops Frobenius twists and Frobenius kernels to reduce questions about quotients in general tensor categories to the classical theory of algebraic groups, and proves that $\mathcal{G}/\mathcal{H}$ exists, is of finite type, and is algebraic/separated; moreover, its geometric type (quasi-affine, affine, or proper) agrees with that of the body quotient $\mathcal{G}_0/\mathcal{H}_0$. The results yield a robust framework for equivariant sheaves and induction, with a detailed treatment in the Ver_p setting where the category is semisimple, allowing explicit presentations of the quotients. The work advances the program of understanding representation theory and geometric structures in incompressible tensor categories, and provides tools for extending classical geometric methods to the tensor-categorical landscape, including applications to Ver_p and related categories. The methodology blends Takian-style Hopf-algebra techniques, fppf-descent formalism, and Frobenius-twisted reductions to connect categorical geometry with traditional algebraic geometry.
Abstract
Let $\mathscr{C}$ be a symmetric tensor category of moderate growth, and let $\mathcal{H}\subseteq\mathcal{G}$ be algebraic groups in $\mathscr{C}$. We prove that the homogeneous space $\mathcal{G}/\mathcal{H}$ exists and is of finite type when $\mathscr{C}$ satisfies (GR) and (MN1-2), which are conjecturally equivalent to incompressibility. A key tool is the introduction of a Frobenius kernel of an group scheme. We further show that while $\mathcal{G}_0/\mathcal{H}_0$ and $(\mathcal{G}/\mathcal{H})_0$ need not be the same, they are close enough, so that $\mathcal{G}/\mathcal{H}$ is quasi-affine/affine/proper if and only if $\mathcal{G}_0/\mathcal{H}_0$ is.
