Recursos
Resumen
Take an odd, absolutely irreducible 2-adic representation with no geometry at 2 at all, and explain how it still ends up as a point of the full completed Hecke algebra T₂(N). The trick is to stop looking at that representation by itself.
The target
The full completed Hecke algebra T₂(N) is an inverse limit over all weights up to k, as k grows [0:28]. Pro-modularity [0:47] means a continuous λ: T₂(N) → 𝒪_E with λ(T_ℓ) = tr r(Frob_ℓ) and λ(ℓS_ℓ) = det r(Frob_ℓ) for ℓ ∤ 2N. That makes r a Hecke point, not a classical modular form. Scalar and reducible residual cases are included.
Three moves
- Family [1:36]. A continuous cyclotomic twist moves Hecke points to Hecke points, so you may normalize det r = χ = δ·εʷ and undo the twist at the end. Deform freely with that determinant fixed. The trace image is a local domain D with dim D ≥ 3, and r is one of its points.
- Regular point [2:13]. Find a point x on D that is de Rham at 2 with distinct Hodge–Tate weights. Here p-adic Langlands is a black box: base change split at 2, Colmez's correspondence, positive Iwasawa rank, then a depth-zero supercuspidal type. Regular Fontaine–Mazur makes x classical up to a Tate twist, so x lies on Hecke support.
- Tangent squeeze [3:02]. Every component of the Hecke quotient A has dimension 4, so dim A_x = 3. Newton–Thorne bounds edim R_x ≤ 3. That gives 3 ≤ dim A_x ≤ dim R_x ≤ edim R_x ≤ 3. Every inequality becomes an equality, R_x is regular, and the Hecke kernel vanishes at x.
Then the result spreads [4:04]. Any kernel element is killed by some s outside x, and s stays nonzero on the domain D. So the kernel vanishes on all of D, including r. This is the algebraic cousin of the identity theorem.
Where the de Rham condition is used [4:39]: only in move 2, and only at the auxiliary point x. It never applies to r itself.