LA02 — Taniyama-Shimura and Fermat’s Last Theorem

Field Value
Domain Langlands Programme
System Elliptic curves over ℚ and modular forms
Group GL_2(ℤ̂) (automorphic side) / Gal(ℚ̄/ℚ) → GL_2(ℤ_ℓ) (Galois side)
H^k tier H²
ISA Meld (β→0)
Status Validated
Opcodes ORBIT · TWIST · BIND · LABEL · FLIP
Papers Paper 492, Paper 469, Paper 543

Physical system

The Taniyama-Shimura-Weil conjecture (proved by Wiles 1995 for semistable elliptic curves, completed by Breuil-Conrad-Diamond-Taylor 2001 for all) states:

Every elliptic curve E over ℚ is modular: there exists a modular form f of weight 2 and level N_E (the conductor of E) such that the L-functions agree:

L(s, E) = L(s, f)

where L(s,E) = ∏_p (1−a_p(E)p^{−s}+p^{1−2s})^{-1} (with a_p(E) = p+1−#E(𝔽_p)) and L(s,f) = ∏_p (1−a_p(f)p^{−s}+p^{k−1−2s})^{-1} (Hecke eigenvalues).

This is the GL_2 case of the Langlands correspondence over ℚ:

{ 2-dimensional Galois representations ρ_{E,ℓ}: Gal → GL_2(ℤ_ℓ) } ↔ { weight-2 newforms f ∈ S_2(Γ_0(N)) }

Fermat’s Last Theorem follows: if aⁿ+bⁿ=cⁿ had a solution with n≥3, Frey (1986) constructed an elliptic curve E_{a,b,c} whose Galois representation would be modular but whose conductor would not satisfy the constraints of any modular form (Ribet’s theorem, 1990). Contradiction. ∎

The ISA reading: FLT is an ORBIT closure argument — the Galois ORBIT of a hypothetical Frey curve cannot close (no matching modular form TWIST), so the Frey curve cannot exist, so the FLT equation has no solutions.

Target category

ModGal(ℚ) — the category of compatible systems of ℓ-adic Galois representations over ℚ, whose objects are continuous homomorphisms ρ_ℓ: Gal(ℚ̄/ℚ) → GL_2(ℤ_ℓ) for all primes ℓ, satisfying compatibility: tr(ρ_ℓ(Frob_p)) ∈ ℤ is independent of ℓ for all p outside a finite set. Morphisms are isomorphisms of representations. The Taniyama-Shimura theorem proves that every elliptic curve’s system {ρ_{E,ℓ}} lies in the modular subcategory.

Interpretation functor

F: C → ModGal(ℚ) defined by:

Opcode F(opcode)
ORBIT Elliptic curve points: E(ℚ̄) = the group of solutions (x,y) to y²=x³+ax+b over ℚ̄; the Galois group Gal acts on E(ℚ̄) by permuting coordinates; the ℓ-adic Tate module Tℓ(E) = lim_{n} E[ℓⁿ] ≅ ℤ_ℓ² is the ORBIT on 2-dimensional ℓ-adic space
TWIST Modular form: f(τ) = Σ a_n q^n (q=e^{2πiτ}) is a weight-2 newform with Fourier coefficients a_p = Hecke eigenvalues; the TWIST is the modular group action γ: τ ↦ (aτ+b)/(cτ+d) on the upper half-plane; f transforms as f(γτ) = (cτ+d)²f(τ) (weight-2 TWIST)
BIND Modularity: the Galois representation ρ_{E,ℓ} ≅ ρ_{f,ℓ} (they are isomorphic as GL_2 representations); this isomorphism is the BIND — the non-Abelian holonomy identifying the two sides; Wiles proved this BIND exists for all semistable E
LABEL a_p(E) = p+1−#E(𝔽_p): the number of points on E mod p; equals the Hecke eigenvalue a_p(f) after BIND; this is the primary LABEL eigenvalue — computable by counting points on E mod p
FLIP Complex conjugation c: Gal → GL_2; c acts on ρ_{E,ℓ} as the matrix [[0,−1],[1,0]] (the Weil pairing FLIP); c acts on the modular form as f(−τ̄) = f̄(τ); the FLIP connects the two real embeddings of ℚ

ISA programme 

CURVE:   LABEL[E: y^2 = x^3 + ax + b | elliptic curve over Q, a,b in Z]
POINTS:  ORBIT[E(Q_bar) | Galois acts on all geometric points]
TATE:    ORBIT[T_ell(E) = Z_ell^2 | 2-dim ell-adic representation]
GALOIS:  ORBIT[rho_{E,ell}: Gal(Q-bar/Q) -> GL_2(Z_ell) | Galois rep]
COUNT:   LABEL[a_p = p+1 - #E(F_p) | point count mod p, prime p]
FORM:    TWIST[f(tau) = sum a_n q^n | modular form, same a_p]
HECKE:   TWIST[T_p f = a_p f | Hecke eigenvalue equation]
BIND:    BIND[rho_{E,ell} iso rho_{f,ell} | Taniyama-Shimura = BIND]
FREY:    ORBIT[E_{a,b,c}: y^2=x(x-a^n)(x+b^n) | hypothetical Frey curve]
RIBET:   BIND[rho_{Frey,ell} not modular | no BIND possible by level-lowering]
FLT:     LABEL[therefore a^n+b^n != c^n for n>=3 | Fermat output]

Computable output

  • Point counts a_p(E) = p+1−#E(𝔽_p): the primary LABEL eigenvalue, computed by enumerating E(𝔽_p) for each prime p. For E: y²=x³−x (conductor N=32): a_2=0, a_3=0, a_5=−2, a_7=0, a_{11}=0, a_{13}=6, … These equal the Hecke eigenvalues of the modular form f=η(2τ)⁴η(4τ)⁴/η(8τ)⁴ (weight 2, level 32). The match a_p(E) = a_p(f) for all p is the BIND verification.

  • Wiles’s R=T theorem (the BIND proof): Wiles proved that the deformation ring R (classifying Galois representations deforming ρ̄{E,ℓ}) is isomorphic to the Hecke algebra T (classifying modular forms with the same Hecke eigenvalues). R ≅ T is the algebraic statement of the BIND ρ{E,ℓ} ≅ ρ_{f,ℓ}. This required proving that R and T have the same order — a numerical coincidence proved using the Euler characteristic of a certain Selmer group (a TWIST at H¹).

  • Selmer group Sel_ℓ(E/ℚ): the H¹ content of LA02; it measures the obstruction to lifting rational points from E(ℚ) to E(ℚ̄). #Sel_ℓ = ℓ^r where r is the “Selmer rank” (upper bound for the Mordell-Weil rank). The Birch–Swinnerton-Dyer conjecture (a separate Millennium problem) relates #Sel_ℓ to the order of vanishing of L(s,E) at s=1 — see the Millennium section below.

  • Fermat’s Last Theorem: for n≥3, no (a,b,c) ∈ ℤ³ with abc≠0 satisfies aⁿ+bⁿ=cⁿ. If such (a,b,c) existed:

    1. Frey curve E_{a,b,c}: y²=x(x−aⁿ)(x+bⁿ) would be semistable (Frey’s construction)
    2. Its Galois representation ρ̄_{E,ℓ} would be modular (Wiles: all semistable curves are modular) with some level N
    3. Ribet (1990): N must divide 2 (level-lowering theorem for Frey curves)
    4. There are no weight-2 newforms of level 2 (direct computation: dim S_2(Γ_0(2))=0)
    5. Contradiction: the BIND ρ_{E,ℓ} ≅ ρ_{f,ℓ} cannot exist (no f to bind to)

    The ISA logic: FLT = an empty ORBIT (no Frey curve exists because no BIND target exists in the modular space).

The β-plane and GL_2 Langlands

The GL_2 Langlands correspondence over ℚ lives at s = 1/2 on the critical line of the adèlic β-plane (Paper 543 §6):

Quantity β-plane location
a_p(E) = LABEL at prime p β_p = 1/√p (local)
L(s,E) = product LABEL s = Re(β) = 1/2 (functional equation)
Modular form f β = iτ (Im(β) = τ in upper half-plane)
Taniyama-Shimura BIND β: Q-adelic → complex β-plane via modularity

The functional equation L(s,E) = ±N^{1−s} L(2−s,E) is a β-plane symmetry: s ↔ 2−s corresponds to β ↔ 1/β (the Origami ↔ Meld duality of the β-plane). The sign ε_E = ±1 (root number) determines whether L(s,E) vanishes at s=1 (even analytic rank) or not (odd analytic rank) — the BSD conjecture is the claim that this ε_E sign encodes the rank of E(ℚ) via a β* snap.

Why H² (not H¹)

  • The ℓ-adic Tate module Tℓ(E) is a 2-dimensional GL_2(ℤ_ℓ) representation — the fundamental 2×2 matrix representation. This is H² because GL_2 is non-Abelian (unlike GL_1 = H¹ of GA01).
  • The BIND ρ_{E,ℓ} ≅ ρ_{f,ℓ} is an isomorphism of non-Abelian representations: it is not just a scalar matching (χ_E = χ_f as in H¹) but a full matrix equivalence (requires checking all GL_2 structure, not just determinant).
  • The R=T theorem (the BIND proof) requires the full derived functor machinery — it is a statement in derived algebraic geometry (derived deformation rings) that cannot be reduced to H¹.

Connections to other entries

  • GA01 (Galois cyclotomic): LA02 generalises GA01 from GL_1 to GL_2; the cyclotomic character χ_ℓ: Gal → ℤ_ℓ× is the GL_1 analogue of ρ_{E,ℓ}; the Dirichlet L-function is the GL_1 analogue of L(s,E)
  • GA02 (FeMoco): the G₂ representation of FeMoco (GA02) is the exceptional- group analogue of ρ_{E,ℓ}; Paper 492 develops the molecular Langlands connection
  • LA01 (Geometric Langlands): LA02 is the number-theoretic version (over ℚ) of LA01 (over 𝔽_q curves); they are unified in the global Langlands programme
  • MT02 (Apéry ζ(3)): ζ(3) is an L-value at s=3 of the Dedekind zeta function; the Apéry recurrence (MT02) = a modular form recurrence at level 6 (Beukers 1987); both are LABEL eigenvalues of Galois representations
  • CP01 (3-SAT): Wiles’s proof required 7 years and multiple error corrections — the longest β-stay in the H² Meld regime in modern mathematics; the proof complexity is H² (requires full derived algebraic geometry, not H¹ techniques)

Validation

  • Wiles (1995), Ann. Math. 141, 443 + Taylor-Wiles (1995), Ann. Math. 141, 553: modularity for semistable elliptic curves over ℚ; Fermat’s Last Theorem proved. Fields Medal 1998.
  • Breuil, Conrad, Diamond & Taylor (2001), J. Am. Math. Soc. 14, 843: full modularity theorem for all elliptic curves over ℚ.
  • Ribet (1990), Invent. Math. 100, 431: level-lowering; Frey curve’s Galois representation has no modular form at level 2.
  • Frey (1986): construction of E_{a,b,c} from a hypothetical FLT solution; the Frey curve trick that converted FLT to modularity.
  • Serre (1987): conjectural ε-conjecture (now Ribet’s theorem); connecting FLT to GL_2 Langlands via level-lowering.

Part of the ISA Zoo. Categorical foundations: Paper 591.