3.2 Local Langlands correspondence
3.2.2 Local Langlands correspondence for GL(2, F )
In this subsection, we remind the reader the theory of irreducible admis- sible representations of GL(2, F ). After that, we attach L-parameters to these representations. In our description of the irreducible admissible rep- resentations of GL(2, F ) we will use the same notation as the one we used when we listed the irreducible admissible non-supercuspidal representations for GSp(4, F ); thus we need the notion of the Grothendieck group (see also Definition 2.3.1). That is, if A is the category of smooth representations of finite length of GL(2, F ), the Grothendieck group of A is the abelian group generated by isomorphism classes π (for simplicity we skip the notation “h” and “i”) of objects in A, modulo the relations
π2 = π1+ π3,
for all short exact sequences π1 ,→ π2 π3 in A.
The first thing to say is that the finite-dimensional irreducible admissible representations of GL(2, F ) are one-dimensional. In fact, they are of the form χ ◦ det for some character χ of F×. This is Proposition 2.7(a) of [32].
We consider now infinite dimensional irreducible admissible representa- tions of GL(2, F ). The minimal parabolic subgroup of GL(2, F ) is the Borel subgroup B consisting of all upper triangular matrices; that is
B = ( t1 x t2 !) .
This is in fact the unique proper parabolic subgroup of GL(2, F ), up to conjugation. If χ1 and χ2 are two characters of F×, define the character
t1 x
t2
!
7→ χ1(t1)χ2(t2)
of the Borel. We apply normalized induction on this character, to get a representation of GL(2, F ) with representation space consisting of all locally
constant functions f on GL(2, F ) such that f ( t1 x t2 ! g) = t1 t2 1/2 χ1(t1)χ2(t2)f (g), for t1 x t2 !
∈ B and g ∈ GL(2, F ). The group GL(2, F ) acts on this space by right translation. We denote this representation by χ1×χ2. The following
theorem lists the non-supercuspidal irreducible admissible representations of GL(2, F ). Note that by Proposition 2.2.2, the representation χ1 × χ2 is
admissible.
Theorem 3.2.7. For the parabolically induced representation χ1 × χ2 of
GL(2, F ), we have the following:
i. The representation χ1× χ2 is irreducible when χ1χ−12 6= | |
±1. In this
case we have χ1× χ2 ∼= χ2× χ1.
ii. If χ1χ−12 = | |, we have that there is an irreducible subrepresentation
denoted by (χ2| |1/2)StGL(2) and an irreducible quotient (χ2| |1/2) ◦ det
of dimension one. Considering these representations as elements in the Grothendieck group, we may write
χ1× χ2 = (χ2| |1/2)StGL(2)+ (χ2| |1/2) ◦ det.
iii. If χ1χ−12 = | |−1, we have that there is an irreducible one-dimensional
subrepresentation (χ1| |1/2) ◦ det and an irreducible quotient denoted by
(χ1| |1/2)StGL(2). Considering these representations as elements in the
Grothendieck group, we write
χ1× χ2 = (χ1| |1/2) ◦ det + (χ1| |1/2)StGL(2).
Proof. This is Theorem 3.3 in [32].
If the representation χ1 × χ2 is irreducible, we say that it is a princi-
resentation of Definition 2.2.3. In order to have all irreducible admissible representations of GL(2, F ), we also need to consider the representations which do not arise as subrepresentations or subquotients of representations obtained by normalized induction from the Borel parabolic; these are the supercuspidal representations which we briefly describe below.
Supercuspidal representations are in general difficult to describe, but if we assume that the residual characteristic p is odd, things become easier. Let L/F be a quadratic extension and ψ an admissible character of L× such that ψ ◦ σ 6= ψ. Here σ is the non-trivial element of Gal(L/F ). Then one gets an irreducible admissible supercuspidal representation of GL(2, F ) from ψ, called a base change. We denote this supercuspidal representation by BC(L/F, ψ). This construction can be found in more detail in Theorem 4.6 of [32]. If p 6= 2 then every supercuspidal representation of GL(2, F ) arises that way. The extra irreducible admissible representations that one gets if p = 2 are called extraordinary representations.
Definition 3.2.8. An L-parameter for the group GL(2, F ) is an equivalence class of admissible representations of the Weil-Deligne group WF0. Denote by Φ(GL(2, F )) the set of L-parameters.
Let us now describe the L-parameters which are attached to the above representations (at least when p is odd). Below, we will be seeing characters of F×as characters of WF and vice versa, without distinguishing the notation.
i. To the principal series representations χ1× χ2, we attach L-parameters
(ρ0, N ), with semisimple part ρ0 : WF → GL(2, C) given by
ρ0 : w 7→
χ1(w)
χ2(w)
! ,
and nilpotent part N = 0.
L-parameter (ρ0, N ), with ρ0 : WF → GL(2, C) defined via ρ0 : w 7→ χ(w)|w| χ(w) ! .
Here the nilpotent part is non-trivial; in fact N = 1 !
.
iii. To a supercuspidal representation BC(L/F, ψ) we attach the L-parameter (ρ0, N ) with trivial nilpotent endomorphism N , and ρ0 = indWWFLψ.
We will usually write φπ for the L-parameter of a representation π of
GL(2, F ).
The local Langlands correspondence for GL(2, F ) has been proved by Kutzko in [36] (and more generally for GL(n, F ) by Harris and Taylor in [30] and Henniart in [31]). The supercuspidal representations are in bijec- tion with irreducible 2-dimensional representations of the Weil-Deligne group (i.e., irreducible representations of the Weil group as we have seen above), the principal series representations correspond with semisimple representations of WF0 which are direct sums of two characters, and the twisted Steinberg repre- sentations correspond with reducible indecomposable representations of WF0. It is now clear that the consideration of twisted Steinberg representations is forcing us to choose the Weil-Deligne group instead of the Weil group in the correspondence. Furthermore, the central characters of irreducible admissi- ble representations are the determinants of the correpsponding Weil-Deligne representations; in particular, we have
ωχ1×χ2 = χ1χ2;
ω(χ| |1/2)St
GL(2) = χ
2| |;
ωBC(L/F,ψ) = ψ|F×L/F.
Here L/F is the quadratic character corresponding to the quadratic extension
L/F ; for x ∈ F× we define L/F(x) = 1 if x is such that there is a x0 ∈ L×
norm map of the extension L/F .
For completeness, we shall discuss briefly the archimedean L-parameters. For this we will consider only the complex archimedean place since this will be the case of interest to us. In this case, the Weil group is WC = C×, and the definition of an archimedean L-parameter is analogous to the non- archimedean case (see Definition 3.2.8). For integers n ≥ 0 and w with n ≡ w + 1 mod 2, define φw,n : WC → GL(2, C) via
z 7→ |z|−w (z/¯z)
n/2
(z/¯z)−n/2 !
.
According to §3.1 of [40], such a representation corresponds to an irreducible admissible representation of GL(2, C) (the latter are described in §6 of [32]).