Elliptic curves and modularity

Let E be an elliptic curve over Q and choose a minimal integral Weierstrass model. At a prime p of good reduction, count its points over Fₚ including the point at infinity and set aₚ=p+1−#E(Fₚ). Hasse's bound gives |aₚ|≤2√p. The good local Euler denominator is 1−aₚp⁻ˢ+p^(1−2s), matching the weight-two Hecke denominator. A finite count is an exact arithmetic computation; a numerical comparison of complex values is not needed to recover that local coefficient.

Bad primes require different factors. For multiplicative reduction, aₚ is plus one for split and minus one for nonsplit reduction, with denominator 1−aₚp⁻ˢ. For additive reduction the local factor is one, conventionally aₚ=0 for this purpose. The conductor N records local reduction data with exponents, not simply the product of primes in the discriminant with their discriminant exponents. An equation's discriminant must first be interpreted in a minimal model before conductor conclusions are drawn.

The modularity theorem states that every elliptic curve E/Q of conductor N has a normalized cuspidal newform f in S₂(Γ₀(N)) with trivial character, integral Fourier coefficients, and L(E,s)=L(f,s), including the bad-prime factors. Normalized means a₁(f)=1, and new means that its genuine arithmetic level is N rather than an oldform inherited from a smaller level. The semistable case was proved by Wiles and Taylor–Wiles; the full theorem was completed by Breuil, Conrad, Diamond, and Taylor. These are deep theorems, not deductions from finitely many matching primes.

Conversely, a normalized weight-two newform with rational Fourier coefficients determines an elliptic curve over Q up to Q-isogeny. Rationality is crucial: a higher-degree coefficient field generally produces a higher-dimensional modular abelian variety instead. Isogenous rational curves share this L-function, so coefficients do not select one Weierstrass equation uniquely. The correspondence naturally concerns an isogeny class. Distinct equations within that class can have different torsion presentations, periods, or degrees of a chosen parametrization without contradicting the common coefficient sequence.

A geometric formulation provides a nonconstant morphism X₀(N)→E defined over Q; one can translate it so that a chosen rational cusp maps to the identity. Analytically, integrating the weight-two differential associated with f produces periods and a complex torus, but descending the construction to Q and matching the conductor are substantive arithmetic steps. Complex uniformization alone applies to every complex elliptic curve and therefore cannot prove this rational modularity statement. A complex parameter τ need not encode an elliptic curve defined over Q.

Modularity gives analytic continuation and a functional equation for the elliptic curve's L-function through its modular form. It does not, by itself, identify the order of its central zero with the rank of rational points. That equality belongs to the Birch–Swinnerton-Dyer conjecture in general, alongside further leading-coefficient predictions. A few checked point counts provide useful evidence and catch normalization errors, but a proof that a proposed form corresponds to a curve must invoke the theorem or an adequate separate identification argument. Good-prime agreement alone should never be described as a proof of every local factor.