Fermat's Last Theorem Blueprint

7. Automorphic forms and the Langlands Conjectures🔗

7.1. Definition of an automorphic form for GLn over Q🔗

The global Langlands reciprocity conjectures relate automorphic forms to Galois representations. The statements for a general connected reductive group involve the construction of the Langlands dual group, and we do not have quite enough Lie algebra theory to push this definition through in general. However if we restrict the special case of the group \GL_n/\Q, the dual group is just \GL_n(\bbC) and several other technical obstructions are also removed. In this section we will explain the definition of an automorphic form for the group \GL_n/\Q, following the exposition by Borel and Jacquet in Corvallis (volume, 1979).

7.2. The Finite Adeles Of The Rationals🔗

Mathlib already has the definition of the finite adeles \A_{\Q}^f of the rationals as a commutative \Q-algebra, and the proof that it's a topological ring.

7.3. The Group GLn Of The Adeles🔗

The adeles \A_{\Q} of \Q are the product \A_{\Q}^f \times \R, with the product topology. They are a topological ring. Hence \GL_n(\A_{\Q}) = \GL_n(\A_{\Q}^f) \times \GL_n(\R) is a topological group, where we are being a bit liberal with our use of the equality symbol.

7.4. Smooth Functions🔗

A function f : \GL_n(\A_{\Q}^f) \times \GL_n(\R) \to \bbC is smooth if it has the following three properties:

  1. f is continuous.

  2. For all x \in \GL_n(\A_{\Q}^f), the function y \mapsto f(x,y) is smooth.

  3. For all y \in \GL_n(\R), the function x \mapsto f(x,y) is locally constant.

Current state of this definition: I've half-formalised it; I don't know how to say the the function is smooth on the infinite part, because I have never used the manifold library before and I have no idea what my model with corners is supposed to be.

7.5. Slowly-Increasing Functions🔗

Automorphic representations satisfy a growth condition which we may as well factor out into a separate definition.

We define the following temporary "size" function s : \GL_n(\R) \to \R by s(M) = \operatorname{trace}(MM^T + M^{-1}M^{-T}), where M^{-T} denotes inverse-transpose. Note that s(M) is always positive, and is large if M has a very large or very small, in absolute value, eigenvalue.

We say that a function f : \GL_n(\R) \to \bbC is slowly-increasing if there is some real constant C and positive integer n such that |f(M)| \leq C s(M)^n for all M \in \GL_n(\R).

Lean code for Definition7.5.11 definition
  • structure(1 field)defined in FLT/GlobalLanglandsConjectures/GLnDefs.lean
    complete
    structure AutomorphicForm.GLn.IsSlowlyIncreasing {n : } (f : GL (Fin n)   ) :
      Prop
    structure AutomorphicForm.GLn.IsSlowlyIncreasing
      {n : } (f : GL (Fin n)   ) : Prop
    A function `f : GL_n(ℝ) → ℂ` is slowly increasing if there exist `C, N` such
    that `‖f M‖ ≤ C · s(M)^N` for all `M`, where `s` is an auxiliary "size" function. 
    bounded_by :  C N,  (M : GL (Fin n) ), f M  C * AutomorphicForm.GLn.s M ^ N

Note: the book says n is positive, but \{M|s(M)\leq 1\} is compact so I don't think it makes any difference.

7.6. Weights At Infinity🔗

The weight of an automorphic form for \GL_n/\Q can be thought of as a finite-dimensional continuous complex representation \rho of a maximal compact subgroup of \GL_n(\R), and it's convenient to choose one (they're all conjugate) so we choose O_n(\R).

Lean code for Definition7.6.11 definition
  • structure(2 fields)defined in FLT/GlobalLanglandsConjectures/GLnDefs.lean
    complete
    structure AutomorphicForm.GLn.Weight (n : ) : Type
    structure AutomorphicForm.GLn.Weight (n : ) : Type
    A weight for `GLₙ`: a preweight whose associated `O(n)`-representation is simple. 
    w : AutomorphicForm.GLn.preweight n
    The underlying preweight, i.e. continuous representation of `O(n)`. 
    isSimple : CategoryTheory.Simple (AutomorphicForm.GLn.preweight.fdRep n self.w)

The Lean definition is incomplete right now -- I don't demand irreducibility (I wasn't sure whether I was doing this the right way; if I used category theory then I might have struggled to say that the representation was continuous).

7.7. The Action Of The Universal Enveloping Algebra🔗

There is a natural action of the real Lie algebra of \GL_n(\R) on the complex vector space of smooth complex-valued functions on \GL_n(\R).

This extends to a natural complex Lie algebra action of the complexification of the real Lie algebra on the smooth complex-valued functions on \GL_n(\R). This depends on Definition 7.7.1.

By functoriality, we get an action of the universal enveloping algebra of this complexified Lie algebra on the smooth complex-valued functions. This depends on Definition 7.7.2.

Thus the centre Z_n of this universal enveloping algebra also acts on the smooth complex-valued functions. This depends on Definition 7.7.3.

The centre we just defined is a commutative ring which contains a copy of \bbC. Note that Harish-Chandra, or possibly this was known earlier, showed that it is a polynomial ring in n variables over the complexes. We shall not need this.

7.8. Automorphic Forms🔗

From here on there is no more Lean right now, only LaTeX.

Definition7.8.1
Statement uses 4
Statement dependency previews
used by 0L∃∀N

A smooth function f : \GL_n(\A_{\Q}^f) \times \GL_n(\R) \to \bbC is an O_n(\R)-automorphic form on \GL_n(\A_{\Q}) if it satisfies the following five conditions. This depends on Definition 7.4.1, Definition 7.5.1, Definition 7.6.1, and Definition 7.7.4.

  1. Periodicity: for all g \in \GL_n(\Q), we have f(gx,gy) = f(x,y).

  2. It has a finite level: there exists a compact open subgroup U \subseteq \GL_n(\A_{\Q}^f) such that f(xu,y) = f(x,y) for all u \in U, x \in \GL_n(\A_{\Q}^f), and y \in \GL_n(\R).

  3. It has weight \rho: there exists a continuous finite-dimensional irreducible complex representation \rho of O_n(\R) such that for every (x,y) \in \GL_n(\A_{\Q}), the complex vector space spanned by the functions k \mapsto f(x,yk) is finite-dimensional and isomorphic, as an O_n(\R)-representation, to a direct sum \rho^{\oplus m} of copies of \rho for some m.

  4. It has an infinite level: there is an ideal I \subseteq Z_n of the centre Z_n described in the previous section, with finite complex codimension, and I annihilates the function y \mapsto f(x,y) for all x \in \GL_n(\A_{\Q}^f). This is a very fancy way of saying that the function satisfies some natural differential equations. In the case of modular forms, these are the Cauchy-Riemann equations, which is why modular forms are holomorphic.

  5. It satisfies the growth condition: for every x \in \GL_n(\A_{\Q}^f), the function y \mapsto f(x,y) on \GL_n(\R) is slowly-increasing.

Lean code for Definition7.8.11 definition
  • structure(5 fields)defined in FLT/GlobalLanglandsConjectures/GLnDefs.lean
    complete
    structure AutomorphicForm.GLn.AutomorphicFormForGLnOverQ (n : )
      (ρ : AutomorphicForm.GLn.Weight n) : Type
    structure AutomorphicForm.GLn.AutomorphicFormForGLnOverQ
      (n : )
      (ρ : AutomorphicForm.GLn.Weight n) :
      Type
    Automorphic forms for GL_n/Q with weight ρ. 
    toFun : GL (Fin n) (IsDedekindDomain.FiniteAdeleRing  ) × GL (Fin n)   
    The underlying function `GL_n(𝔸_f) × GL_n(ℝ) → ℂ`. 
    is_smooth : AutomorphicForm.GLn.IsSmooth self.toFun
    is_periodic :  (g : GL (Fin n) ) (x : GL (Fin n) (IsDedekindDomain.FiniteAdeleRing  )) (y : GL (Fin n) ),
      self.toFun
          (((algebraMap  (IsDedekindDomain.FiniteAdeleRing  )).GL (Fin n)) g * x, ((algebraMap  ).GL (Fin n)) g * y) =
        self.toFun (x, y)
    is_slowly_increasing :  (x : GL (Fin n) (IsDedekindDomain.FiniteAdeleRing  )),
      AutomorphicForm.GLn.IsSlowlyIncreasing fun y  self.toFun (x, y)
    has_finite_level :  U, AutomorphicForm.GLn.IsConstantOn U self.toFun

Automorphic forms of a fixed weight \rho form a complex vector space, and if we also fix the finite level U and the infinite level I then we get a subspace which is finite-dimensional; this is a theorem of Harish-Chandra. There is also the concept of a cusp form, meaning an automorphic form for which furthermore some adelic integrals vanish.

7.9. Hecke Operators🔗

The group \GL_n(\A_{\Q}^f) acts on the space of automorphic forms for \GL_n(\A_{\Q}) by the formula (g \cdot f)(x,y) = f(xg,y).

Proof for Lemma 7.9.1
uses 0

This is obvious. Note that the conjugate of a compact open subgroup is still compact and open.

A formal development of the theory of Hecke operators looks like the following.

Let U be a fixed compact open subgroup of \GL_n(\A_{\Q}^f), and let us also fix a weight \rho. Let M_\rho(n) denote the complex vector space of automorphic forms for \GL_n/\Q of weight \rho. The level U forms M_\rho(n,U) are just the U-invariants of this space. If g \in \GL_n(\A_{\Q}^f), then the double coset space UgU can be written as a finite disjoint union of single cosets g_iU: the double coset space is certainly a disjoint union of left cosets, but the double coset space is compact and the left cosets are open.

Define the Hecke operator T_g : M_\rho(n,U) \to M_\rho(n,U) by T_g(f) = \sum g_i \cdot f.

This function is well-defined, i.e. it sends a U-invariant form to a U-invariant form which is independent of the choice of g_i.

Proof for Lemma 7.9.2
uses 0

Easy group theory.