In mathematics, a Drinfeld module (or elliptic module) is roughly a special kind of module over a ring of functions on a curve over a finite field, generalizing the Carlitz module. Loosely speaking, they provide a function field analogue of complex multiplication theory. A shtuka (also called F-sheaf or chtouca) is a sort of generalization of a Drinfeld module, consisting roughly of a vector bundle over a curve, together with some extra structure identifying a "Frobenius twist" of the bundle with a "modification" of it.
Drinfeld modules were introduced by, who used them to prove the Langlands conjectures for GL2 of an algebraic function field in some special cases. He later invented shtukas and used shtukas of rank 2 to provethe remaining cases of the Langlands conjectures for GL2. Laurent Lafforgue proved the Langlands conjectures for GLn of a function field by studying the moduli stack of shtukas of rank n.
"Shtuka" is a Russian word штука meaning "a single copy", which comes from the German noun “Stück”, meaning “piece, item, or unit". In Russian, the word "shtuka" is also used in slang for a thing with known properties, but having no name in a speaker's mind.
We let
L
p>0
L\{\tau\}
a0+a1\tau+a
2+ … | |
2\tau |
L
\taua=ap\tau, a\inL.
The element
\tau
L
L\{\tau\}
L
\tau
L
L\{\tau\}
a0x+a
p2 | |
2x |
+ … =
0+a | |
a | |
1\tau+a |
2+ … | |
2\tau |
in
L[x]
f
f(x+y)=f(x)+f(y)
L[x,y]
L
\tau=xp
infty
infty
infty
Fq[t]
\iota:A\toL
A Drinfeld A-module over L is a ring homomorphism
\phi:A\toL\{\tau\}
\phi
d:L\{\tau\}\toL,a0+a1\tau+ … \mapstoa0
\iota:A\toL
The condition that the image of A is not in L is a non-degeneracy condition, put in to eliminate trivial cases, while the condition that
d\circ\phi=\iota
\iota
As L can be thought of as endomorphisms of the additive group of L, a Drinfeld A-module can be regarded as an action of A on the additive group of L, or in other words as an A-module whose underlying additive group is the additive group of L.
Suppose that X is a curve over the finite field Fp.A (right) shtuka of rank r over a scheme (or stack) U is given by the following data:
E → E′ ← (Fr×1)*E,whose cokernels are supported on certain graphs of morphisms from U to X (called the zero and pole of the shtuka, and usually denoted by 0 and ∞), and are locally free of rank 1 on their supports. Here (Fr×1)*E is the pullback of E by the Frobenius endomorphism of U.
A left shtuka is defined in the same way except that the direction of the morphisms is reversed. If the pole and zero of the shtuka are disjoint then left shtukas and right shtukas are essentially the same.
By varying U, we get an algebraic stack Shtukar of shtukas of rank r, a "universal" shtuka over Shtukar×X and a morphism (∞,0) from Shtukar to X×X which is smooth and of relative dimension 2r - 2. The stack Shtukar is not of finite type for r > 1.
Drinfeld modules are in some sense special kinds of shtukas. (This is not at all obvious from the definitions.) More precisely, Drinfeld showed how to construct a shtuka from a Drinfeld module.See Drinfeld, V. G. Commutative subrings of certain noncommutative rings. Funkcional. Anal. i Prilovzen. 11 (1977), no. 1, 11–14, 96. for details.
See main article: article and Lafforgue's theorem.
The Langlands conjectures for function fields state (very roughly) that there is a bijection between cuspidal automorphic representations of GLn and certain representations of a Galois group. Drinfeld used Drinfeld modules to prove some special cases of the Langlands conjectures, and later proved the full Langlands conjectures for GL2 by generalizing Drinfeld modules to shtukas.The "hard" part of proving these conjectures is to construct Galois representations with certain properties, and Drinfeld constructed the necessary Galois representations by finding them inside the l-adic cohomology of certain moduli spaces of rank 2 shtukas.
Drinfeld suggested that moduli spaces of shtukas of rank r could be used in a similar way to prove the Langlands conjectures for GLr; the formidable technical problems involved in carrying out this program were solved by Lafforgue after many years of effort.