Learning path
From the Riemann hypothesis down to school maths
The outline runs from the two headline theorems through the main steps of the proof, then to background concepts and school or first-year university material. It follows the proof as we read it from the Lean code.
- The Riemann hypothesis theorem
Every zero of the zeta function, other than the "trivial" zeros -2, -4, -6, ..., has real part exactly 1/2. The statement used is Mathlib's own `RiemannHypothesis`, not one written by the project.
On the map: Explicit formula, Perron kernel and zero-free conditions for Dirichlet L-functions Bernoulli numbers and Dirichlet characters
- What RH says concept
ζ(s) starts as 1 + 1/2^s + 1/3^s + ..., which converges for Re s > 1, and is extended to all s ≠ 1. RH is a statement about where that extended function is zero.
- ζ(s) as a series concept
ζ(s) = Σ 1/n^s for complex s with Re s > 1, where the series converges.
- Analytic continuation concept
A complex-differentiable function on a region has at most one extension to a bigger region. This is how ζ gets a meaning where its series diverges.
- Trivial zeros −2, −4, … concept
ζ is zero at every negative even integer. In the functional equation this comes from the cosine factor; these zeros are "trivial" and excluded from RH.
- Functional equation of ζ concept
ζ(1-s) = 2 (2π)^(-s) Γ(s) cos(πs/2) ζ(s): values at s and 1-s are tied together, so zeros come in mirror pairs about Re s = 1/2. Proved in Mathlib, not in this project. Using it needs only algebra; proving it needs the theta function and Poisson summation.
- The Gamma function concept
Γ(s) = ∫₀^∞ t^(s-1) e^(-t) dt extends the factorial (Γ(n+1) = n!) to complex s.
- Fourier analysis, Poisson summation concept
Writing functions as sums of waves. Poisson summation relates the sum of f(n) over integers to the sum of its Fourier transform; it is the engine behind the functional equation.
- L2 for all characters ⇒ RH step
Apply L2 to the trivial character mod 1, whose L-function is ζ itself. That gives "ζ has no zeros right of 1/2", and step-rh-of-right finishes.
On the map: Bernoulli numbers and Dirichlet characters
- No zeros right of ½ ⇒ RH step
A symmetry argument. If s is a zero with Re s < 1/2, the functional equation ζ(s) = 2 (2π)^(s-1) Γ(1-s) cos(π(1-s)/2) ζ(1-s) forces ζ(1-s) = 0, because the other factors are nonzero (the cosine vanishes only at trivial zeros). But Re(1-s) > 1/2, a contradiction. So every non-trivial zero has Re s = 1/2. About 60 lines, mostly edge cases (s = 0, signs of k).
On the map: Bernoulli numbers and Dirichlet characters
- L2: no zeros right of ½, every Dirichlet L-function step
For every N, every Dirichlet character χ mod N, and every s with Re s > 1/2: L(χ, s) ≠ 0 (except the pole of ζ at s = 1). This is the real content of the proof; nearly all 175,000 files exist to prove it.
On the map: Host-field conditions, inertness of primes and character box parameters
- Dirichlet L-functions concept
L(χ, s) = Σ χ(n)/n^s, ζ twisted by a character. ζ is the case N = 1.
- L2 from literature + "VWin" step
Guess from the name: L2 follows from a package of results quoted from the literature, plus a condition the code calls "VWin".
- Local Langlands existence (recTExists) step
Guess from the name: an existence statement in the local Langlands correspondence.
- Galois local–global compatibility step
Guess from the name: a strong local–global compatibility statement for Galois representations.
- "VWin" from a modularity argument step
Guess from the folder name: a modularity argument that supplies "VWin".
- GRH for Dirichlet L-functions theorem
The same for Dirichlet L-functions: every zero in the strip 0 < Re s < 1 has real part 1/2. Stated in Statements.lean, which imports only Mathlib.
On the map: Explicit formula, Perron kernel and zero-free conditions for Dirichlet L-functions Bernoulli numbers and Dirichlet characters
- L2 ⇒ GRH (mirror trick for L-functions) step
The same mirror argument for a primitive character χ. A zero s of L(χ, s) with Re s < 1/2 gives a zero 1-s of L(χ̄, s) with Re > 1/2, which L2 rules out. The mirror swaps χ with its conjugate, so L2 must hold for all characters at once.
On the map: Bernoulli numbers and Dirichlet characters
- Functional equation of L(χ, s) concept
For primitive χ, the completed function Λ(χ, s) (L times a Gamma factor that depends on whether χ is even or odd) satisfies Λ(χ, 1-s) = ε Λ(χ̄, s). Proved in Mathlib.