Fermat's Last Theorem: 358 Years From Margin Note to Proof

A Margin Too Small for History

Around 1637, Pierre de Fermat scribbled a note in the margin of his copy of Diophantus' Arithmetica: "I have discovered a truly remarkable proof of this theorem, but this margin is too small to contain it." The theorem itself was simple to state: no three positive integers a, b, c satisfy aⁿ + bⁿ = cⁿ for any integer n greater than 2. The proof Fermat claimed to possess was almost certainly wrong — the tools required to actually prove it would not exist for another 350 years.

The Problem That Resisted Everyone

Fermat's Last Theorem (FLT) became the most famous unsolved problem in mathematics — not because it was necessarily the deepest, but because its statement was comprehensible to anyone who had learned basic algebra. That accessibility attracted an unusual range of attempted proofs: professional mathematicians, amateur enthusiasts, and crackpots all submitted arguments to journals and prize committees over three centuries.

Concrete progress came in fragments. In the late 17th century, Fermat himself proved the n = 4 case using a technique called infinite descent. Leonhard Euler proved the n = 3 case in 1770 (though his original proof had a gap, later corrected). Sophie Germain developed a general approach in the early 19th century that proved FLT for a class of prime exponents now called Germain primes. By 1993, the theorem had been verified computationally for all n up to 4 million — but a computational check is not a proof.

The Bridge: Elliptic Curves and Modular Forms

The route to a complete proof ran through two seemingly unrelated areas of mathematics.

Elliptic Curves

An elliptic curve is an equation of the form y² = x³ + ax + b, studied over the rational numbers or over finite fields. These curves have a remarkably rich structure — their solutions form a group, and classifying elliptic curves is a central problem in number theory. In 1985, Gerhard Frey made a critical observation: if Fermat's Last Theorem were false — if there existed integers a, b, c, n with aⁿ + bⁿ = cⁿ — then one could construct an elliptic curve from those numbers that would have very strange properties.

Modular Forms

Modular forms are complex analytic functions with extraordinary symmetry properties — they transform in predictable ways under a group of transformations of the upper half-plane. Every modular form has an associated L-function, a type of generating series encoding arithmetic information. Modular forms are the territory of serious analytic number theory.

The Shimura-Taniyama-Weil Conjecture

In 1955, Yutaka Taniyama posed a question connecting these two worlds. Goro Shimura developed the idea further, and André Weil added precision in 1967: the conjecture states that every elliptic curve over the rational numbers is modular — that is, associated to a modular form in a specific, precise sense. This "Shimura-Taniyama-Weil conjecture" (now called the Modularity Theorem) was the bridge.

In 1986, Ken Ribet proved that if the Shimura-Taniyama-Weil conjecture were true, then Fermat's Last Theorem must also be true — because Frey's hypothetical curve could not be modular, contradicting the conjecture. Proving FLT now required proving that every elliptic curve is modular. It was an enormous problem, but at least it was a well-defined one.

Andrew Wiles: Seven Years in Secret

Andrew Wiles encountered Fermat's Last Theorem as a ten-year-old in a Cambridge library in 1963. When Ribet's result appeared in 1986, Wiles — by then a professor at Princeton — recognized that a career mathematician's ambition had become technically accessible. He did not tell colleagues what he was working on. He worked in his attic, avoiding the usual cycle of presentations and preprints that would telegraph his approach to competitors.

The Gap and the Rescue

The 1993 announcement turned into crisis when Nick Katz, checking the proof, found a problem in the "Euler system" approach Wiles had used to control Selmer groups — a technical mechanism essential to the argument. For more than a year, Wiles struggled to repair the gap.

On September 19, 1994, in a moment Wiles later described as the most important of his professional life, he saw how to combine two failed approaches — Iwasawa theory and the Euler system — in a way that made each repair the defect of the other. The insight resolved the gap completely. The final proof runs to 109 pages in the Annals of Mathematics, with an additional 45-page supporting paper by Taylor and Wiles.

What the Proof Actually Contains

The Wiles proof does not directly attack the equation aⁿ + bⁿ = cⁿ. Instead, it proves a substantial case of the Shimura-Taniyama-Weil conjecture (for semistable elliptic curves), which forces the conclusion via Ribet's work. The core tools are:

• Galois representations: Encoding the symmetries of elliptic curves into representations of the absolute Galois group of the rationals
• Hecke algebras: Algebras of operators acting on spaces of modular forms
• Deformation theory: Studying how Galois representations vary in families
• Iwasawa theory: A technique from algebraic number theory controlling the size of Selmer groups

Fermat's proof — if he had one — could not have used any of these tools, which were invented centuries after his death. Most mathematicians believe Fermat had found a flawed argument for the n = 4 case that he incorrectly believed generalized.

Legacy and Open Questions

Wiles received the Abel Prize in 2016, which many consider the closest thing mathematics has to a Nobel Prize. The proof's most significant legacy is not FLT itself but the Modularity Theorem it required — the complete proof (for all elliptic curves, not just semistable ones) was established by Breuil, Conrad, Diamond, and Taylor in 2001, with enormous implications for algebraic number theory well beyond the theorem that prompted it.

• The Generalized Fermat equation — x^p + y^q = z^r for distinct prime exponents — remains largely open
• The ABC conjecture, which would generalize FLT, is claimed to have been proved by Shinichi Mochizuki but has not achieved consensus acceptance as of 2024
• The Langlands Program — a vast network of conjectures connecting number theory and representation theory — is the direct descendant of the ideas in Wiles' proof