Posed 1859 · Hilbert's 8th problem (1900) · Clay Millennium Prize (2000, $1,000,000) · open 167 years
How to read this page. We use: s = σ + it (a complex number, real part σ, imaginary part t); ζ(s) = the Riemann zeta function; Re(s) = real part; Σ = sum, ∏ = product over primes p; Γ = the gamma function (factorial, extended); li(x) = the logarithmic integral; π(x) = the count of primes ≤ x. If any symbol stops you, the Understanding Math & Symbols primer on the home page defines all of them.
0 · The claim, in one line
Every "non-trivial" zero of the zeta function lies exactly on the vertical line Re(s) = ½. Not approximately. Not almost all. Every single one, out to infinity.
1 · The object: and why anyone cares
Start with an innocent-looking sum (Euler, 1730s), valid when Re(s) > 1:
ζ(s) = Σn=1∞ 1/ns = 1 + 1/2s + 1/3s + 1/4s + ⋯
Euler's miracle: this sum equals a product running over only the prime numbers:
ζ(s) = ∏p prime 1 / (1 − p−s)This single identity is the bridge. Because it holds, every fact about ζ(s) is secretly a fact about the primes.
That equality is the whole reason the hypothesis matters. The zeta function is a dictionary that translates the chaotic distribution of primes into the smooth language of complex analysis. Control the zeros of ζ, and you control the error in every estimate of how primes are spread out.
2 · Extending the function: analytic continuation
The sum above diverges for Re(s) ≤ 1. Riemann (1859) showed ζ can be uniquely extended to the whole complex plane (one pole at s = 1) obeying a mirror symmetry: the functional equation. Define the completed function:
That symmetry s ↔ 1−s reflects the plane across the line Re(s) = ½. The critical line is not an arbitrary choice: it is the axis of symmetry of the function itself. That is the first hint the zeros might live there.
3 · The zeros
Trivial zeros: at s = −2, −4, −6, … (forced by the Γ(s/2) factor). Fully understood, uninteresting.
Non-trivial zeros: all lie in the critical strip0 < Re(s) < 1. These encode the primes. The hypothesis says every one has Re(s) = ½.
The critical strip (shaded, 0 < Re < 1), the critical line Re(s)=½ (dashed), the first two trivial zeros (✕ at −2, −4), and the first five non-trivial zeros: all sitting exactly on the line, at heights t ≈ 14.13, 21.02, 25.01, 30.42, 32.94.
4 · Every connection: what the zeros secretly control
The zeros are not a curiosity. Through Riemann's explicit formula, they literally reconstruct the primes. Let ψ(x) count prime powers up to x; then
ψ(x) = x − Σρ xρ/ρ − ln(2π) − ½ ln(1 − x−2)The sum is over all non-trivial zeros ρ. Each zero contributes a wave of amplitude set by its real part. If every ρ has Re = ½, every wave has amplitude √x: the smallest possible, giving the tightest control on the primes.
This is why RH is equivalent to a whole constellation of other statements. A proof of any one proves them all:
Prime counting: RH ⟺ π(x) = li(x) + O(√x · ln x): the best-possible error term in the Prime Number Theorem.
Möbius / Mertens: RH ⟺ M(x) = Σn≤x μ(n) = O(x½+ε). Honesty note: the stronger "Mertens conjecture" |M(x)| < √x was disproven (Odlyzko & te Riele, 1985): a warning that intuitive strengthenings of RH can be flat wrong.
Robin's inequality (1984): RH ⟺ σ(n) < eγ n ln ln n for all n > 5040 (σ = sum of divisors). A purely elementary restatement: no complex analysis in sight.
Li's criterion (1997): RH ⟺ a specific sequence λn ≥ 0 for every n.
Nyman–Beurling: RH ⟺ a certain set of functions is dense in a Hilbert space: recasts RH as an approximation problem.
5 · Every serious attack: and exactly how far each got
(a) Put zeros on the line directly
Hardy (1914): proved infinitely many zeros lie on the line. (But "infinitely many" is not "all": infinitely many could still be off it.)
Selberg (1942): a positive proportion lie on the line. Levinson (1974): at least ⅓. Conrey (1989): at least ⅖ (40.77%); nudged past 41% since.
(b) Compute them
The first ten trillion (1013) non-trivial zeros have been calculated (Gourdon–Demichel, 2004): all on the line. Spot-checks near heights of 1022–1032 (Odlyzko, Platt): still on the line.
(c) Find a hidden structure that forces the line
Hilbert–Pólya program: find a self-adjoint operator (a quantum Hamiltonian) whose energy levels are the zeros' heights. Self-adjoint operators have real eigenvalues → zeros forced onto the line.
Montgomery–Dyson (1973): the spacing statistics of the zeros match the eigenvalue statistics of random Hermitian matrices (the GUE ensemble) exactly. Overwhelming evidence a Hilbert–Pólya operator exists: but no one has built it. Berry–Keating guessed the operator is a quantized H = xp; it reproduces the statistics but has never been made rigorous.
The function-field analogue: PROVEN. Replace the integers with polynomials over a finite field. There, the exact analogue of RH is a theorem: Weil (1948) for curves, Deligne (1974) for the general case. This is the single most important clue: and §6 is why it does not transfer.
6 · Why it breaks down: the honest crux
Every route above hits a wall. Here is precisely where each one fails, which is the part most pages skip:
The proven case can't be ported. Weil and Deligne proved the function-field RH using geometry that the integers do not have. Over a finite field there is an actual curve (a geometric object) and the zeros become eigenvalues of the Frobenius map acting on its cohomology; a positivity theorem (Weil positivity / the Riemann–Roch theorem on the surface) forces them onto the circle. For the ordinary integers, Spec ℤ, there is no known curve, no cohomology, and no positivity theorem. Fifty years of trying to manufacture one (Arakelov geometry, Connes' noncommutative spaces, the conjectural "field with one element" 𝔽₁) have produced beautiful frameworks but not the one positivity statement that would close the proof.
The "positive proportion" ceiling is real. Levinson/Conrey work by multiplying ζ by a cleverly chosen "mollifier" and counting sign changes. The length of the mollifier is capped by what we can control, and that cap mathematically limits the proportion to well under 100%. Pushing to 100% is not a matter of effort: the method has no mechanism that could ever reach all zeros. It proves "most," never "every."
No operator has ever been exhibited. Hilbert–Pólya is a hope, not a construction. The GUE match is stunning evidence, but statistics of a spectrum do not hand you the operator, and Berry–Keating's H = xp has no confining boundary condition, so its spectrum isn't even discrete without extra ad-hoc surgery. Sixty years: still no self-adjoint operator.
Computation is logically incapable of proving it. There are infinitely many zeros; a single counterexample could sit arbitrarily far out. History screams caution here: π(x) − li(x) is negative for every x anyone could ever compute, yet it must flip sign: the first crossing is near 10316 (the Skewes phenomenon). A property true for the first 1013 cases, or 10300, is no proof at all for an infinite claim.
Zero-free regions crawl. The classical approach (de la Vallée Poussin, 1896) proves a thin zero-free region hugging the edge Re(s) = 1, widened slightly since. But the known techniques widen it only logarithmically: they provably cannot reach all the way in to the line Re(s) = ½. The gap is not shrinking fast enough to ever close.
The pattern. Every attack falls into exactly one of three traps: (1) it proves a strictly weaker statement with a hard mathematical ceiling (positive proportion, zero-free strip); (2) it needs a structure that exists for polynomials but demonstrably not for the integers (the curve, the cohomology, the operator); or (3) it is empirical and cannot, even in principle, settle an infinite statement. No known idea escapes all three traps at once. That intersection (weaker-but-capped, structure-that-doesn't-exist, or logically-insufficient) is where the Riemann Hypothesis has lived, unbroken, for 167 years.
7 · Status: confirmed
Unsolved Verified against the historical and mathematical record. No valid proof or disproof exists. All ~1013 computed zeros are consistent with RH, but consistency is not proof.
What a genuine proof would require: either (i) construct the Hilbert–Pólya operator explicitly and prove it self-adjoint; or (ii) build the missing geometry/positivity over Spec ℤ that mirrors Weil–Deligne; or (iii) a wholly new idea belonging to none of the three traps above. Anything less (more computed zeros, a higher proportion, a wider zero-free region) strengthens belief without settling the question.
Credit where it is due. Riemann (1859, the hypothesis & explicit formula); Euler (the product); Hadamard & de la Vallée Poussin (1896, PNT & zero-free region); Hardy (1914); Selberg, Levinson, Conrey (proportion on the line); Montgomery & Dyson (1973, the random-matrix connection); Weil (1948) & Deligne (1974, the proven function-field case); Odlyzko & te Riele (1985, disproof of the Mertens conjecture); Robin (1984) & Li (1997, equivalences); Gourdon–Demichel, Platt, Odlyzko (computation). The unsolved problem itself is claimed by no one here: if it is ever solved, that proof is offered freely to the world.