We reach the summit — arguably the single most famous unsolved problem in all of mathematics. The Clay Mathematics Institute has a standing one-million-dollar prize sitting unclaimed for whoever proves it, and mathematicians have been trying since 1859.

The Riemann Hypothesis is a single, crisp claim about where the zeta function equals zero. That sounds narrow — a question about one function's zeros — but it is really a question about the primes: it says exactly how evenly they are scattered among the ordinary counting numbers, and pins down the tightest possible bound on how far the prime-counting function can ever stray from its smooth prediction. Prove it, and you don't just win the money — you sharpen one of the deepest results in number theory at a single stroke.

The two kinds of zeros

The analytically continued \zeta(s) has zeros of two kinds. The trivial zeros sit at the negative even integers s = -2, -4, -6, \dots and are completely understood — they fall straight out of the functional equation that ties \zeta(s) to \zeta(1-s):

Worked example. Set s = -2. The factor \sin(\pi s/2) = \sin(-\pi) = 0, so the whole right-hand side collapses to zero — hence \zeta(-2) = 0. The same trick kills \sin(\pi s /2) at every negative even integer, which is exactly why the trivial zeros sit where they do. Nobody worries about these; they're accounted for.

The non-trivial zeros are the interesting ones. All of them lie in the critical strip 0 < \Re(s) < 1 — that much has been proved — and it is their exact horizontal position inside that strip that the whole hypothesis is about.

The hypothesis

In words: take any zero that isn't one of the "trivial" ones above, write it as s = \sigma + it, and the claim is that \sigma is always exactly 1/2 — never 0.499, never 0.5001, always dead-centre. It is one of the seven original Clay Millennium Prize Problems.

Picturing the claim

It helps to see the strip itself. Below, the horizontal axis is the real part \Re(s) and the vertical axis is the imaginary part \Im(s). Step through: first the strip that must contain every non-trivial zero, then the single line down its middle, then the actual heights at which the first known zeros occur — every one of them landing precisely on that line.

Worked example. The very first non-trivial zero (going up from the real axis) sits at s \approx 0.5 + 14.1347i. Its real part is exactly 1/2, just as the hypothesis demands. The next few sit at imaginary parts near 21.02, 25.01, 30.42, and 32.94 — again, every single one exactly on the critical line.

What's actually known

Here is the strange part: the evidence for the Riemann Hypothesis is overwhelming. Riemann himself checked the first few zeros by hand in 1859. In 1953, Alan Turing — yes, that Turing — built one of the earliest computer programs specifically to hunt for zeros off the line, and found none. Since then, computers have verified more than 10^{13} (ten trillion) non-trivial zeros directly, and separate, cleverer arguments (counting how many zeros must exist in a range, then checking that many really do sit on the line) confirm that no zero has been skipped over within the ranges searched.

And yet: not one single non-trivial zero has ever, in over 165 years of searching, been found anywhere off the critical line. That is about as strong as numerical evidence in mathematics ever gets — and it is still not a proof.

The bulk of that ten-trillion-zero search relies on the Riemann–Siegel formula, a way of evaluating \zeta(1/2 + it) for huge t far more efficiently than summing the original series. Refinements of it (and of the computer hardware behind it) let mathematicians such as Andrew Odlyzko push the search deeper and deeper into the critical line through the late twentieth century and beyond — always finding the same thing: every zero, exactly on the line.

Why it matters so much

The location of the zeros controls the error term in the Prime Number Theorem. The Riemann Hypothesis is exactly the statement that the primes are as regularly distributed as they possibly could be — that \pi(x) hugs \operatorname{Li}(x) with the smallest conceivable error:

Worked example. Take x = 10^6. The true count is \pi(10^6) = 78{,}498, and the smooth estimate is \operatorname{Li}(10^6) \approx 78{,}627.5 — a gap of only about 130. Compare that to the RH-style bound, which for this x works out to roughly \sqrt{10^6}\cdot\ln(10^6) = 1000 \times 13.8 \approx 13{,}800. The actual gap is far inside that ceiling, exactly as a true upper bound should be — the real payoff of proving the hypothesis is a guaranteed ceiling for every x, however large, not just a good result for the ones we've already checked.

Hundreds of theorems across mathematics begin "assume the Riemann Hypothesis…" — including results that would guarantee certain primality tests run fast on every input, not just the ones anyone has happened to try. A generalised version of the hypothesis, applied to a broader family of functions related to \zeta, sits quietly underneath a handful of results in computational number theory that in turn touch cryptographic assumptions — one more reason so many mathematicians would love to see it settled, one way or the other.

It is the keystone the whole arch of analytic number theory leans on: known to hold for every zero we can reach, and waiting, still, for a proof that it holds for every zero there is.

See it explained