riemann

The Lean code claims a proof of the Riemann Hypothesis. This site has not checked the claim yet. See what is checked.

The proof in one page

The Lean code proves the Riemann Hypothesis from one statement about Dirichlet L-functions. This page shows that step in words, and shows where the large part of the proof starts.

draft

The claim

The Lean code states the Riemann Hypothesis with the definition of Mathlib, and gives a proof:

theorem riemannHypothesis : RiemannHypothesis := rhOfAll

The definition RiemannHypothesis of Mathlib says this. Let ss be a complex number with ζ(s)=0\zeta(s) = 0. If ss is not a trivial zero −2,−4,−6,…-2, -4, -6, \dots and s≠1s \neq 1, then Re⁡(s)=12\operatorname{Re}(s) = \tfrac12. The statement is on the line Main.lean#L4. The words of this site do not prove it. They tell what the Lean code does. The trust report tells what this site has checked.

The chain at the top

The top of the proof is short. It has four steps, and each step is a theorem in the Lean code.

  1. riemannHypothesis is rhOfAll (MainTheorems.lean#L9).
  2. rhOfAll applies rh_of_every to allChi. Here allChi is the statement EveryCharacter below.
  3. rh_of_every (CritLine.lean#L29) takes EveryCharacter for the trivial character modulo 1. The L-function of that character is the Riemann zeta function. Thus ζ(s)≠0\zeta(s) \neq 0 when Re⁡(s)>12\operatorname{Re}(s) > \tfrac12 and s≠1s \neq 1.
  4. rh_of_right (RhOfRight.lean#L5) gets the Riemann Hypothesis from that half-plane. It uses only Mathlib.

Step 4 in words

Suppose that ζ(s)=0\zeta(s) = 0, that ss is not a trivial zero and that s≠1s \neq 1. There are three cases.

  • If Re⁡(s)>12\operatorname{Re}(s) > \tfrac12, step 3 says that ζ(s)≠0\zeta(s) \neq 0. This is not possible.
  • If Re⁡(s)=12\operatorname{Re}(s) = \tfrac12, the claim is true.
  • If Re⁡(s)<12\operatorname{Re}(s) < \tfrac12, use the functional equation of Mathlib, riemannZeta_one_sub:
ζ(1−s)=2(2π)−(1−s) Γ(1−s)cos⁡ ⁣(π(1−s)2)ζ(s).\zeta(1 - s) = 2 (2\pi)^{-(1-s)} \, \Gamma(1 - s) \cos\!\left(\frac{\pi (1 - s)}{2}\right) \zeta(s) .

Read it at the point 1−s1 - s, so that the left side is ζ(s)=0\zeta(s) = 0. On the right side, Re⁡(1−s)>12\operatorname{Re}(1 - s) > \tfrac12, so ζ(1−s)≠0\zeta(1 - s) \neq 0 by step 3. The factor 2(2π)−(1−s)2(2\pi)^{-(1-s)} is not zero. The Gamma function has no zeros, so Γ(1−s)≠0\Gamma(1 - s) \neq 0. The cosine is zero only when s=−2ks = -2k for an integer kk. If k>0k > 0, then ss is a trivial zero. If k=0k = 0, then s=0s = 0, but ζ(0)=−12\zeta(0) = -\tfrac12. If k<0k < 0, then s≥2s \geq 2, but Re⁡(s)<12\operatorname{Re}(s) < \tfrac12. Thus the right side is not zero, and this is a contradiction.

So each zero that is not trivial is on the line Re⁡(s)=12\operatorname{Re}(s) = \tfrac12. This part of the proof is classical. A reader who knows the functional equation can check it in a few minutes.

The statement that does the work

All the difficulty is in one statement. The Lean code calls it EveryCharacter (EveryCharacter.lean#L13):

def UpperTheory.EveryCharacter : Prop :=
  ∀ (N : ℕ) [NeZero N] (χ : DirichletCharacter ℂ N) (s : ℂ),
    1 / 2 < s.re → (χ ≠ 1 ∨ s ≠ 1) → DirichletCharacter.LFunction χ s ≠ 0

In words: for each modulus N≥1N \geq 1 and each Dirichlet character χ\chi modulo NN, the Dirichlet L-function L(s,χ)L(s, \chi) has no zero with Re⁡(s)>12\operatorname{Re}(s) > \tfrac12. The pole of ζ\zeta at s=1s = 1 is the only exception. This is the Generalized Riemann Hypothesis for Dirichlet L-functions, in the form "no zeros to the right of the line". Step 20 of the timeline is "GRH", and this is the reason.

The theorem everyCharacter (MainConclusion.lean#L10) proves EveryCharacter. It is one line:

theorem UpperTheory.FinalTheorems.everyCharacter : UpperTheory.EveryCharacter :=
  UpperTheory.FinalTheorems.everyCharacter_of_row UpperTheory.Gal.Int.galoisLocalGlobalStrong_recParameter

Below this line is the large part of the proof: more than 175,000 Lean files on theta lifts, automorphic forms, Galois representations and the local Langlands correspondence. The theorem every_character_of_literature_of_vWin takes a structure of three statements that the Lean code calls OpenLiteratureStatements, and the proof gives each of them from theorems of the Lean code. Milestone M1 makes the map of this part, and the chapters of milestone M3 and later explain it.

Other results in Main.lean

The same file states four more results. They come from the same statement.

Lean name Statement
riemannZeta_eq_zero_re the Riemann Hypothesis, in a second form (L6)
LFunction_ne_zero_of_half_lt_re EveryCharacter itself (L8)
LFunction_eq_zero_re each zero of L(s,χ)L(s, \chi) with 0<Re⁡(s)<10 < \operatorname{Re}(s) < 1 is on the line Re⁡(s)=12\operatorname{Re}(s) = \tfrac12 (L10)
LFunction_eq_zero_re_of_isPrimitive the same for a primitive character, through its functional equation (L12)

The statements are in Statements.lean and use only Mathlib.

What this page does not say

This page does not say that the proof is correct. Lean accepts a proof only relative to the axioms that the proof uses, and a person must check that the Lean statement says what the standard statement says. The trust report answers three questions: is the source the same as the release, which axioms does the proof use, and does a second kernel accept it. Until the report says yes, read each sentence of this site as "the Lean code claims".

The source of this page. Markdown. Something is wrong? Report it.