Recursos
Resumen
Read the geometric prime switch that proves every elliptic curve over every imaginary quadratic field is modular. For each link in the chain, say which prime it is congruent modulo and why, then follow automorphy from an auxiliary curve to E.
The statement and the gap
For any imaginary quadratic field K and any elliptic curve E/K, there is an automorphic representation π_E of GL₂(𝔸_K) with r_ℓ(π_E) ≃ H¹(E). The local parameters match at every place, and L(E/K,s) = L(π_E, s−½). Before this paper, Caraiani–Newton's lifting theorem needed residual hypotheses in characteristic 3 or 5. That left gaps, for example fields where X₀(15)(K) has positive rank [0:37].
The prime switch, built from geometry
Take a non-CM E, a large prime p with mod-p image containing SL₂(𝔽_p), n = 3p and F = K(μ_n). A genus-n cyclic cover of a twist B_t splits under rotation into rank-two pieces V_θ that each look like the cohomology of an elliptic curve. The trivial piece is B_t and an order-3 piece is a new elliptic curve A [1:20].
The chain is V_η ≡ (mod p) V_χ ≡ (mod 3) V_ψ ≡ (mod p) V_1 [2:20]. The rule: a root of unity of ℓ-power order reduces to 1 in characteristic ℓ, so the order-p part of a character dies mod p and the order-3 part dies mod 3. Freeness of the integral cohomology over the group ring turns equal residual characters into equal residual representations, even though p divides n [2:33]. Try the middle link yourself at [3:10].
Moving automorphy
A is modular using the prime 5 (ζ₅ ∉ F). Three lifts follow: at p, then at 3, then at p again, which reaches B over F. You then descend through cyclic prime-degree steps to K and untwist to get E [3:40]. Each lift needs absolute irreducibility. Vanishing cycles supply two opposite unipotents with entry 2 − λ − λ⁻¹ ≠ 0, and no line is fixed by both [4:24].
When you read a prime switch, ask two questions: which part of the character dies modulo which prime, and what keeps the residual image large [5:05].