Let $A$ be a commutative algebra, say over $\mathbb{C}$.

Giving a grading on $A$ corresponds at least morally to giving a $\mathbb{C}^*$ action on spec(A): $A_i$ can be thought of as those functions on which $t$ acts by multiplication with $t^i$. Similary a graded $A$ module is just a $\mathbb{C}^*$ equivariant sheaf.

Now I want to know, if there is also a geometric interpretation of filtered rings/modules.


To a filtered algebra $(A,F)$ one can assign its Rees algebra $R=\bigoplus_i F_iA$. It is a graded algebra containing the algebra of polynomials in one variable $\mathbb{C}[t]$ naturally embedded as the subalgebra generated by the element $t\in R_1$ corresponding to the element $1\in F_1A$. So the algebra $R$ defines a $\mathbb{C}^*$-equivariant quasi-coherent sheaf of algebras $\mathcal{R}$ over the affine line $\operatorname{Spec}\mathbb{C}[t]$. The algebra $A$ can be recovered as the fiber of $\mathcal{R}$ at the point $t=1$, and the associated graded algebra $\operatorname{gr}_FA$ is the fiber of $\mathcal{R}$ at $t=0$. Filtered $A$-modules correspond to $\mathbb{C}^*$-equivariant quasi-coherent sheaves of modules over $\mathcal{R}$.

The algebra $R$ is a torsion-free $\mathbb{C}[t]$-module, so the quasi-coherent sheaf $\mathcal{R}$ over $\operatorname{Spec}\mathbb{C}[t]$ has to be torsion-free. This description does not take into accout the issue of completeness of the filtration $F$ (in case it extends also in the decreasing direction), which requires a separate consideration.

  • $\begingroup$ Thanks, this is exactly the kind of answer I was looking for! $\endgroup$ – Jan Weidner May 25 '10 at 15:26
  • $\begingroup$ In order for 1 to be in F_1 A, we need all F_m to equal R for m nonnegative, and for the OP's A_i to be Leonid's F_{-i}. Then the fiber over 0 is indeed gr A, supported in negative degrees. Anyway, to answer the question "What is the filtered ring A?" the Rees construction answers "A deformation of the graded ring gr A" / "Something that degenerates to the graded ring gr A". $\endgroup$ – Allen Knutson May 27 '10 at 0:12
  • 2
    $\begingroup$ No, one does not need $F_m=A$ for $m$ nonnegative at all. My answer simply presumes the notation in which $F_{i-1} A$ is contained in $F_i A$, and of course 1 is in $F_0 A$, so 1 is in $F_1 A$, to. E.g., the filtration can be increasing, with $F_iA=0$ for $i<0$, or it can be decreasing, with $F_iA=A$ for $i\geq0$, or it can extend nontrivially in both directions. All these cases are covered. $\endgroup$ – Leonid Positselski May 27 '10 at 9:54
  • $\begingroup$ This question and answer is an old favorite of mine. Maybe it's worth adding that if $\mathcal R$ is the Rees algebra of a filtered ring $R$. then a filtered module over $R$ is a $\mathbb G_m$-equivariant quasicoherent sheaf over $Spec \mathcal R$, perhaps with some torsionfreeness condition. $\endgroup$ – Tim Campion May 16 '19 at 21:29

Your Answer

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

Not the answer you're looking for? Browse other questions tagged or ask your own question.