Let ${\mathbb F}:({\mathcal A},{\mathcal E}_{\mathcal A})\to({\mathcal B},{\mathcal E}_{\mathcal B})$ be an exact functor between exact categories, and suppose ${\mathbb F}$ has both a left adjoint ${\mathbb F}_\lambda$ and a right adjoint ${\mathbb F}_\rho$. Then the class

${\mathcal E} := \{X\to Y\to Z\in{\mathcal E}_{\mathcal A}\ |\ {\mathbb F}X\to{\mathbb F}Y\to{\mathbb F}Z\in{\mathcal E}_{\mathcal B}\}$

defines another exact structure on ${\mathcal A}$.

It is interesting to ask for criteria to decide when this exact structure is Frobenius. One such criterion is the following:

Suppose ${\mathcal E}_{\mathcal B}$ is the split exact structure, and that for each $X\in{\mathcal A}$ the unit $\eta_X: X\to{\mathbb F}_\rho{\mathbb F}X$ is an ${\mathcal E}_{\mathcal A}$-monomorphism, while the counit ${\mathbb F}_\lambda{\mathbb F}X\to X$ is an ${\mathcal E}_{\mathcal A}$-epimorphism. Then $({\mathcal A},{\mathcal E})$ has enough projectives and injectives, and the classes ${\mathcal P}$/${\mathcal I}$ of projective/injective objects in $({\mathcal A},{\mathcal E})$ are given by

${\mathcal P} = \{X\in{\mathcal A}\ |\ X\text{ is a summand of some }{\mathbb F}_\lambda , Y\in{\mathcal B}\}$ and

${\mathcal I} = \{X\in{\mathcal A}\ |\ X\text{ is a summand of some }{\mathbb F}_\rho Y, Y\in{\mathcal B}\}$,

respectively. Consequently, if ${\mathbb F}_\lambda$ and ${\mathbb F}_\rho$ have the same image, then $({\mathcal A},{\mathcal E})$ is Frobenius.

This seems very restrictive, but in fact there are at least two cases I know where it can be applied:

(1) If ${\mathcal A}$ is a dg-category, then the forgetful functor ${\mathbb F}: \text{dg-mod}({\mathcal A})\to\text{gr-mod}({\mathcal A})$ fulfills the requirements of the criterion above and thus can be used to construct a Frobenius structure on $\text{dg-mod}({\mathcal A})$ (for pretriangulated dg-categories ${\mathcal A}$, this structure can in turn be restricted to ${\mathcal A}$ itself).

(2) If $G$ is a finite group, $H$ is a subgroup, then the fortgetful functor $G\text{-mod}\to H\text{-mod}$ has left adjoint $\text{Ind}^G_H$ and right adjoint $\text{Coind}^G_H$, and these two functors coincide for $(G:H)<\infty$. In this case, the above criterion therefore applies to provide $G\text{-mod}$ with a Frobenius structure "relative to $H$".


Do you know more criteria for constructing Frobenius structures and situations where they can be applied?

For example, I would be interested in a criterion which can be applied to show that the category of maximal Cohen-Macaulay modules over a Gorenstein ring is Frobenius.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.