MathLabs

Worked solution: Wiles's modularity proof via the Taylor–Wiles method (1994)

Step 4 of 11: Modular forms and modularity explained plainly
In plain words

An elliptic curve EE lives in algebra: for each prime qq where EE is smooth, you can count how many points (x,y)(x, y) it has modulo qq and record the error aq(E)=q+1−∣E(Fq)∣a_q(E) = q + 1 - |E(\mathbb{F}_q)|. That gives an infinite sequence of integers, one per prime.

A modular form f(τ)f(\tau) lives in a completely different world—complex analysis on the upper half-plane Im⁡(τ)>0\operatorname{Im}(\tau) > 0, where it satisfies rigid 2D-like symmetries under fractional linear transformations τ↦(aτ+b)/(cτ+d)\tau \mapsto (a\tau + b)/(c\tau + d) and expands as a Fourier series ∑n≥1c(n,f)qτn\sum_{n \ge 1} c(n, f) q_{\tau}^n. Saying that EE is modular means there is a modular form ff whose Fourier coefficients c(q,f)c(q, f) match the point-count errors aq(E)a_q(E) for every good prime qq.

aq(E)=q+1−∣E(Fq)∣=c(q,f),f(τ)=∑n=1∞c(n,f) e2πinτa_q(E) = q + 1 - |E(\mathbb{F}_q)| = c(q, f), \quad f(\tau) = \sum_{n=1}^{\infty} c(n, f)\, e^{2\pi i n \tau}
Detailed analysis

For an elliptic curve E/QE/\mathbb{Q} and a prime q∤Δq \nmid \Delta of good reduction, let ∣E(Fq)∣|E(\mathbb{F}_q)| be the number of solutions (x,y)∈Fq2(x, y) \in \mathbb{F}_q^2 plus the point at infinity O\mathcal{O}, and define the trace of Frobenius aq(E)=q+1−∣E(Fq)∣a_q(E) = q + 1 - |E(\mathbb{F}_q)| (Qiu et al. 2025, Section 3.2.1). These integers package into the Hasse–Weil LL-series L(E,s)=∑n=1∞an(E)n−sL(E, s) = \sum_{n=1}^{\infty} a_n(E) n^{-s}.

On the analytic side, a cusp form ff of weight k=2k = 2 and level NN for the congruence subgroup Γ0(N)={(a bc d)∈SL⁡2(Z):N∣c}\Gamma_0(N) = \{\binom{a\ b}{c\ d} \in \operatorname{SL}_2(\mathbb{Z}) : N \mid c\} is a holomorphic function on the upper half-plane satisfying f((aτ+b)/(cτ+d))=(cτ+d)2f(τ)f((a\tau+b)/(c\tau+d)) = (c\tau+d)^2 f(\tau) for all matrices in Γ0(N)\Gamma_0(N) and vanishing at the cusps (Wiles 1995, p. 445; Qiu et al. 2025, Section 3.2.2). When ff is an eigenform for all Hecke operators TnT_n, say Tnf=c(n,f)fT_n f = c(n, f) f, its Fourier expansion f(τ)=∑n=1∞c(n,f)e2πinτf(\tau) = \sum_{n=1}^{\infty} c(n, f) e^{2\pi i n \tau} carries arithmetic eigenvalues c(n,f)c(n, f).

By the theorem of Eichler and Shimura (Wiles 1995, p. 444–445, Theorem 0.1), an elliptic curve E/QE/\mathbb{Q} of conductor NN is modular—meaning it is covered by the modular curve X0(N)X_0(N)—if and only if there is a weight-22, level-NN Hecke eigenform ff with rational coefficients such that c(q,f)=aq(E)c(q, f) = a_q(E) for every prime q∤Nq \nmid N. This bridges the world of cubic curves with the rigid finite-dimensional spaces of modular forms.

Terms in this step
Cusp form of weight 22 and level NN
A holomorphic function f(τ)f(\tau) on the complex upper half-plane that transforms by (cτ+d)2(c\tau+d)^2 under matrices (a bc d)∈Γ0(N)\binom{a\ b}{c\ d} \in \Gamma_0(N) with N∣cN \mid c and vanishes at all cusps. For each fixed level NN, these functions form a finite-dimensional vector space.
Hecke eigenform
A modular form ff that is a simultaneous eigenvector for all Hecke averaging operators TnT_n, so Tnf=c(n,f)fT_n f = c(n, f) f. Its Fourier coefficients c(n,f)c(n, f) are algebraic integers that satisfy multiplicative relations.
Knowledge used in this step