There is the following version of categorical local geometric class field theory:

Let $\mathbb{D}=\operatorname{Spec} \mathbb{C}((t))$, $L\mathbb{G}_m$: the loop group of $\mathbb{G}_m$ over $\mathbb{C}$ (i.e. $L\mathbb{G}_m(A)=\mathbb{G}_m(A((t)))$ for any $\mathbb{C}$-algebra $A$), then there is an equivalence of triangulated categories: $\operatorname{Dmod}(L\mathbb{G}_m) \simeq \operatorname{QCoh}(\operatorname{Locsys}_{\mathbb{G}_m}\mathbb{D})$, where the first category is the D-modules on the loop group and the second category is the quasi-coherent sheaves on the stack of rank 1 local systems on $\mathbb{D}$.

I have seen this "theorem" in least two places. One is the paper Tate's thesis in the de Rham setting of Hilburn and Raskin in which they claimed the equivalence is due to Beilinson–Drinfeld. Another place is a talk of Braverman in which he claimed the equivalence is due to Laumon. Both places however did not seem to give a detailed reference (e.g. [BD, Thm x.xx] or [Lau, Cor x.xx]). I believe a proof has to do with certain Fourier–Mukai transform, but I'd really like to see the details if they were written down. Can someone point me to a reference for the proof?

  • $\begingroup$ MathSciNet does not know a paper by Hilburn and Raskin. What is the title? $\endgroup$
    – LSpice
    Apr 1, 2022 at 1:16
  • $\begingroup$ @LSpice: Tate's thesis in the de Rham Setting $\endgroup$
    – wkf
    Apr 1, 2022 at 1:20
  • $\begingroup$ Thanks! I have edited in the name. $\endgroup$
    – LSpice
    Apr 1, 2022 at 1:53
  • 3
    $\begingroup$ In your post, should the definition of $L\mathbb{G}_m(A)$ be the invertible elements in $A((t))$? Or are you considering the loop space of the addtive group $\mathbb{G}_a$? $\endgroup$ Apr 1, 2022 at 10:10
  • 2
    $\begingroup$ @JasonStarr, absolutely, thanks for pointing out the typo. I corrected it. $\endgroup$
    – wkf
    Apr 1, 2022 at 18:42

1 Answer 1


As pointed out in the comments, this is Theorem of Hilburn-Raskin. (It certainly was known much earlier, but I'm not sure what to give as a reference.)

Their proof is stated quite elegantly, in a way that makes the canonicity of the equivalence clear. Let me give a less canonical explanation, which you may or may not find more digestible. (But note that if you want to show good properties of this equivalence it really does help to have a canonical description.)

The point is that $L\mathbb{G}_m$ and $\operatorname{LocSys}_{\mathbb{G}_m}(\mathbb{D})$ are relatively simple and can be described explicitly.

Let's start with $L\mathbb{G}_m$. Actually, let's start with its reduced (ind-)scheme $(L\mathbb{G}_m)_{\operatorname{red}}.$ By Kashiwara's lemma, this has the same category of D-modules as does $L\mathbb{G}_m.$ I claim there is an isomorphism


where $\mathbb{A}^{\infty}_{\operatorname{pro}}$ is $\operatorname{Spec}$ of a polynomial ring in infinitely many generators. The -pro subscript reflects that this is a scheme which is an inverse limit of finite-dimensional affine spaces; later we will see its cousin $\mathbb{A}^{\infty}_{\operatorname{ind}}$ which is the ind-scheme that is a colimit of finite-dimensional affine spaces.

Let me justify this isomorphism on $\mathbb{C}$-points and leave verifying it at the level of functors of points as an exercise. The $\mathbb{C}$-points of $L\mathbb{G}_m$ parametrize invertible elements of $\mathbb{C}((t))$. There are of the form


with $d\in\mathbb{Z}$, $c$ an invertible complex number, and $a_i$ complex numbers. The $d$ corresponds to the $\mathbb{Z}$-factor, the $c$ gives the $\mathbb{G}_m$, and the $a_i$ give the $\mathbb{A}^{\infty}_{\operatorname{pro}}.$

Now we turn to $\operatorname{LocSys}_{\mathbb{G}_m}(\mathbb{D}).$ I claim there is an isomorphism

$$\operatorname{LocSys}_{\mathbb{G}_m}(\mathbb{D})\cong B\mathbb{G}_m\times(\mathbb{A}^1/\mathbb{Z})\times(\mathbb{A}^{\infty}_{\operatorname{ind}})_{\operatorname{dR}}.$$

Here $B\mathbb{G}_m$ is the classifying stack of $\mathbb{G}_m$ and $\mathbb{A}^1/\mathbb{Z}$ is the quotient of the affine line by the translation $\mathbb{Z}$-action. The last term is a little more subtle. As mentioned before, there is an ind-scheme $\mathbb{A}^{\infty}_{\operatorname{ind}}$ defined as a colimit of finite dimensional affine spaces. I take its de Rham stack, which has functor of points


The main property I will use of de Rham stacks is that $\operatorname{QCoh}((\mathbb{A}^{\infty}_{\operatorname{ind}})_{\operatorname{dR}})\cong\operatorname{D-mod}(\mathbb{A}^{\infty}_{\operatorname{ind}})$.

Once again, let me check this expression for $\operatorname{LocSys}_{\mathbb{G}_m}(\mathbb{D})$ at the level of $\mathbb{C}$-points. Of course, you should be fundamentally unsatisfied with this, given that the de Rham stack is invisible if you only look at $\mathbb{C}$-points, but I leave the functor of points check as an exercise. The moduli space of local systems is the quotient


where $\mathbb{C}((t))^{\times}$ acts by gauge transformations. Because we are in an abelian setting, the gauge action is super simple; the action of $f$ is just translation by $\frac{df}{f}.$ Again, write


Then we have


The $c$ does not appear in the final expression, indicating that the $\mathbb{G}_m$ corresponding to $c$ acts trivial, which contributes a $B\mathbb{G}_m$. The $\operatorname{dlog}(1+a_1t+a_2t^2+\cdots)$ can be any formal power series, so all in all the $\mathbb{C}$-points of our quotient are elements of


The coefficient of $\frac{1}{t}$ gives a $\mathbb{A}^1/\mathbb{Z}$, and the coefficients of $\frac{1}{t^n}$, $n\geq 2$, give a $(\mathbb{A}^{\infty}_{\operatorname{ind}})_{\operatorname{dR}}.$ To see that we get $\mathbb{A}^{\infty}_{\operatorname{ind}}$ and not $\mathbb{A}^{\infty}_{\operatorname{pro}},$ note that only finitely many coordinates are allowed to be nonzero. (If you want to see where the de Rham stack comes from, you'll have to do this calculation at the level of functors of points.) So that gives our expression for $\operatorname{LocSys}_{\mathbb{G}_m}(\mathbb{D}).$

Once we have these formulas, we can see that geometric local class field theory is the product of three simpler equivalences:




The first equivalence follows because $\operatorname{QCoh}(B\mathbb{G}_m)$ is the representation category of $\mathbb{G}_m$, which can be described as the category of $\mathbb{Z}$-graded vector spaces.

The second equivalence is the Fourier transform for D-modules on $\mathbb{G}_m$ (it sends the D-module $Dt^{\lambda}$ on $\mathbb{G}_m$ to $\lambda$.)

The third equivalence is a limit of finite-dimensional additive Fourier transforms of D-modules.

Combined they give the desired equivalence.


You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .