Level 5
Where We Stand
165 years of attack: what we know, what we believe.
You now understand the statement — genuinely, mechanically, the way Riemann meant it. This last level is the situation report: what 165 years of assault have produced, what they haven’t, and what it is reasonable to believe. Click through the campaign:
Bernhard Riemann's only paper on number theory, 'On the Number of Primes Less Than a Given Magnitude,' continues ζ to the complex plane, proves the functional equation, sketches the explicit formula — and remarks in passing that all nontrivial zeros 'very probably' lie on the critical line. He calls his own attempts to prove it 'fleeting' and 'futile,' and moves on.
What the computers know
The first 1013 nontrivial zeros — ten trillion of them — have been checked, every one sitting exactly on the critical line. (Our own little precompute script repeated the first hundred of those checks to build this site; the data files match the published tables to nine decimal places.) So the problem is settled, surely?
No — and you have already seen why. Level 2’s Skewes’ number showed that π(x) − Li(x) flips sign infinitely often, while every direct computation in history sits inside the region where it never flips: the first crossover may lie past 10300. The Mertens conjecture was “verified” to billions and is false. In number theory, the integers we can reach are not a fair sample of the integers that exist. Quantities like ln ln x grow so slowly that behaviour relevant to RH may only emerge at heights no computation will ever visit.
Yet honesty cuts both ways: the computations are not worthless. They have falsification power — one off-line zero would end RH, and ten trillion attempts have failed to find one. They feed the statistics below, and they certify RH-dependent results in explicit ranges. Belief in RH today rests on a tripod: unbroken computation, partial theorems (40% of zeros provably on the line, almost all provably near it), and structural coincidences like the one that follows.
The physics rhyme
The strangest evidence is statistical. Properly rescaled, the gaps between consecutive zeros follow — to extraordinary precision — the distribution physicists derived for energy levels of heavy atomic nuclei (the Gaussian Unitary Ensemble of random matrix theory).
Bars: how often two of our first 100 zeros sit at (rescaled) distance r from each other. Gold curve: the random-matrix (GUE) prediction Montgomery and Dyson stumbled on in 1973. Note the dip near r = 0 — zeros avoid crowding, exactly as quantum energy levels do. With only 100 zeros the histogram is coarse; Odlyzko’s computations with millions of zeros around the 10²⁰th match this curve almost perfectly.
Eigenvalues of certain operators are automaticallyreal — and “all zeros on the critical line” is, after a change of coordinates, precisely the statement that certain numbers are all real. Hence the Hilbert–Pólya conjecture: the zeros are the spectrum of some yet-undiscovered operator, and RH is true because that operator exists. The pair-correlation match is exactly what that picture predicts. It is a rhyme, not a proof — but it is why many experts expect the eventual proof to come from an unexpected direction: spectral theory, mathematical physics, or somewhere stranger.
The bounty
In 2000, the Clay Mathematics Institute named RH one of its seven Millennium Prize Problems: one million dollars for a proof ora disproof. (One of the seven — the Poincaré conjecture — has since fallen, to Grigori Perelman, who declined the money.) The prize is symbolic; the real bounty is the mathematics a proof would have to invent. RH also already appeared on Hilbert’s famous 1900 list, making it the only problem to headline both centuries’ agendas. Hilbert is said to have remarked that if he woke after a five-hundred-year sleep, his first question would be: has the Riemann Hypothesis been proven?
Go deeperThe generalized RH: one hypothesis, a family of functions★★★
ζ has siblings — Dirichlet L-functions, built by twisting the same sum with periodic patterns; they govern primes in arithmetic progressions (primes ending in 1 vs 3 vs 7 vs 9, say). The generalized Riemann Hypothesis (GRH) asserts all their nontrivial zeros also sit on their critical lines. Much of modern number theory is conditional on GRH rather than RH alone — e.g. the deterministic Miller primality test runs in polynomial time under GRH, and effective bounds for the least prime in a progression follow. A proof idea that handles only ζ and not its family would be regarded with suspicion; the truth, whatever it is, should be structural.
Go deeperFamous attempts, and why the problem resists★★★
Approaches that bore other fruit: Weil proved the exact analogue of RH for zeta functions of curves over finite fields (1948), and Deligne for higher dimensions (1974, Fields Medal) — in those worlds, RH is a theorem, proved via a cohomological structure. Nobody has found the analogous structure for the integers, but its existence elsewhere is a standing hint. Connes has recast RH in noncommutative geometry; de Branges’s functional-analytic program has been repeatedly examined and found wanting in detail; countless claimed proofs arrive yearly and dissolve under review.
Why is it hard? The honest short answer: ζ couples addition (the sum over n) to multiplication (the product over p), and mathematics still lacks tools that control both structures at once at this depth. Methods strong enough to certify 40% of zeros provably cannot reach 100% — their own error terms forbid it. And the Rodgers–Tao result says RH, if true, is true with zero slack: no proof can win by a soft, lossy estimate. Whatever works must be exact.
Go deeperThe morning after a proof★★★
Suppose a correct proof appears. What changes overnight? Less than headlines would say: hundreds of conditional theorems become unconditional, prime error terms get certified, and cryptography is untouched (Level 4). What changes over decades could be immense — if the proof builds the long-sought bridge (a spectral operator, a new geometry for the integers), that bridge would carry traffic for a century. A disproof— one stray zero — would be the bigger earthquake: the primes would harbour a conspiracy, a rogue wave at scale x^β with β > ½, and analytic number theory would need rebuilding around it. Almost nobody expects this. The zeros have been interrogated ten trillion times and never once broken formation.
Look again
At the start of this journey you saw the Ulam spiral and were told the diagonals were a mystery. Look at it once more — you see differently now.
Write 1 in the middle, spiral the integers outward, light up the primes. Nobody arranged those diagonal streaks — they emerge. Diagonals correspond to quadratic formulas like n² + n + 41, some of which are unusually rich in primes. The streaks whisper that primes have structure. What structure? That question is this whole site.
The dots are still where they always were. But you now know that behind this scatter stands a smooth law (Level 2); that the law’s fingerprints live in a single function on the complex plane (Level 3); that the scatter itself decomposes into waves, one per zero, like sound into tones (Level 4); and that the deepest open question in mathematics asks only this: are all those waves equally quiet?
Riemann thought it “very probable.” A century and a half of mathematics, and ten trillion zeros, agree with him. Proof is another matter — perhaps for a reader who started, one quiet afternoon, by stepping through a sieve.
↩ Back to Level 1 — or browse the glossary.
What you now know
RH stands unproven but not unmoved: 40% of zeros provably on the line, almost all provably near it, ten trillion checked without exception, the de Bruijn–Newman constant pinned at the boundary, and the zeros’ statistics matching quantum chaos too well for coincidence. You also know why the computations can never settle it — Skewes and Mertens taught you the reach of evidence in this subject. You have climbed from “what is a prime?” to the live edge of mathematics. The mystery is intact. That was the point.