Do All the Zeros Lie on the Critical Line?

The core insight

Whether every nontrivial zero of the Riemann zeta function has real part exactly 1/2 remains unproven — and this single question controls how precisely we can ever know the distribution of the primes.

The question

State it as precisely as the object allows. The Riemann zeta function, for complex numbers $s$ with real part greater than 1, is defined by the convergent series

\zeta(s) = \sum_{n=1}^{\infty} \frac{1}{n^s}.

Riemann showed this function extends, via analytic continuation, to a single-valued holomorphic function on the entire complex plane except for a simple pole at $s=1$ . This extended function has zeros. Some are "trivial": at $s = -2, -4, -6, \dots$ , forced by a symmetry (a functional equation) relating $\zeta(s)$ to $\zeta(1-s)$ . The interesting zeros — the nontrivial ones — all lie in the "critical strip" $0 < \mathrm{Re}(s) < 1$ . This containment is proven; it is not in question.

The Riemann Hypothesis asserts something sharper: every nontrivial zero has real part exactly $1/2$ , lying precisely on the "critical line" $\mathrm{Re}(s) = 1/2$ .

That's the whole statement. Its difficulty lies in what stands behind it. A zero of $\zeta(s)$ is not an isolated curiosity — it is a Fourier-like frequency in the fluctuation of the primes around their expected density. The Hypothesis is really a claim about error terms: how far the count of primes up to $x$ can stray from the smooth prediction given by the prime number theorem. If RH is true, that error is as small as it could possibly be, given the general shape of the problem. If RH is false — if even one zero strays off the line — then somewhere, at some possibly astronomically large scale, the primes clump or thin out in a way fundamentally more irregular than we believe.

What would an answer look like? A proof would almost certainly not be a clever finite calculation; ten trillion confirmed zeros already tell us empirical checking will never suffice, since the claim is about infinitely many zeros. It would need to be a structural argument — quite possibly one connecting zeta's zeros to the spectrum of some operator, or to deep symmetries from a totally different branch of mathematics, in the way that the proof of the analogous statement over finite fields (the Weil conjectures) came from deep algebraic geometry that Riemann never could have imagined. A disproof would be almost anticlimactic by comparison: exhibit one zero, or an argument that infinitely many zeros must eventually drift, and the hypothesis falls, though its shockwaves would rearrange large parts of number theory built provisionally atop it.

What we know

To feel why the zeros matter, you have to see where zeta comes from and what it's secretly counting.

Everything starts with Euler. In the 18th century he noticed a stunning identity: for real $s > 1$ ,

\sum_{n=1}^{\infty} \frac{1}{n^s} = \prod_{p \text{ prime}} \frac{1}{1 - p^{-s}}.

This is the Euler product, and it is really just the analytic shadow of unique factorization: every integer is uniquely a product of prime powers, so summing $1/n^s$ over all integers is the same as multiplying, over each prime $p$ , the geometric series $1 + p^{-s} + p^{-2s} + \cdots$ . The identity converts an additive object — a sum over all integers — into a multiplicative one, a product over the primes alone. It is the first hint that this function, later called $\zeta(s)$ , might encode everything worth knowing about how primes are built and how they're spaced.

Riemann's 1859 memoir — his only paper in number theory, eight pages long — took this further than Euler could have imagined. Euler's product only converges for real $s>1$ ; Riemann realized that treating $s$ as a complex number and extending $\zeta$ by analytic continuation into the rest of the plane reveals structure invisible on the real line alone. Complex analysis is unforgiving and generous at once: a holomorphic function is so rigid that its values on any open set determine it everywhere it can be extended, and singularities and zeros become the load-bearing features of the whole edifice. Riemann showed $\zeta(s)$ continues to a meromorphic function on all of $\mathbb{C}$ , with a single pole at $s=1$ (reflecting the divergence of the harmonic series), and a functional equation

\pi^{-s/2}\,\Gamma(s/2)\,\zeta(s) \;=\; \pi^{-(1-s)/2}\,\Gamma\!\left(\frac{1-s}{2}\right)\zeta(1-s),

which pairs the behavior at $s$ with the behavior at $1-s$ , making the critical strip symmetric about the line $\mathrm{Re}(s) = 1/2$ .

Riemann's real insight, though, was to connect the zeros directly to prime-counting. He derived an exact formula — later made fully rigorous by von Mangoldt — expressing $\pi(x)$ , the count of primes up to $x$ , as a smooth main term (essentially $\mathrm{Li}(x) = \int_2^x dt/\ln t$ ) minus a correction built from a sum over the nontrivial zeros. Each zero $\rho = \beta + i\gamma$ contributes an oscillating term of size roughly $x^\beta$ . The real part $\beta$ of a zero directly controls the amplitude of that oscillation as a function of $x$ . If every $\beta$ equals $1/2$ , every correction term is bounded by $\sqrt{x}$ (times logarithmic factors) — beautifully small, and uniform. If some $\beta$ crept up toward 1, that zero's contribution would grow like $x^\beta$ , swamping the prediction and producing wild, unpredictable clustering or thinning of primes at the corresponding scale.

This is why the Riemann Hypothesis is not really "a statement about a function" in some abstract sense — it is a precise claim about how good the prime number theorem can possibly be. The Prime Number Theorem itself, that $\pi(x) \sim x/\ln x$ , was proven in 1896 by Hadamard and de la Vallée Poussin, using exactly the fact that $\zeta(s) \neq 0$ on the line $\mathrm{Re}(s)=1$ — a weaker, but crucial, zero-free result. RH would upgrade this to the sharpest conceivable form of the theorem: an error term of size $O(\sqrt{x}\log x)$ , matching (up to log factors) the general square-root law you'd expect from a "random" sequence of comparable density, an intuition memorably called the Cramér model.

Somewhere in the strip of complex numbers where the real part sits between 0 and 1, there hide infinitely many points where a certain function vanishes. Ten trillion of these points have been located, computed, checked, and every last one of them has landed on a single vertical line, precisely at real part one-half. No one has ever found an exception. And yet no one has proven there isn't one, lurking at zero number ten-to-the-thirtieth, or ten-to-the-three-hundredth, patiently waiting to break a pattern that has held since Bernhard Riemann first wrote it down in 1859.

This is the Riemann Hypothesis, and it is possibly the most stared-at unproven statement in mathematics. It has survived a century and a half of assault by some of the sharpest minds in history — Hadamard, Hardy, Littlewood, Selberg, Deligne, Bombieri, Sarnak — and it has emerged not just unbroken but strengthened, corroborated by a mountain of computation, echoed in bizarrely distant corners of physics, yet still utterly, maddeningly open. A million-dollar Clay Institute prize sits on the table for whoever settles it. Nobody has come close enough to cash it.

What makes this problem grip people is not just its difficulty but its depth. The zeta function's zeros are not some abstruse curiosity; they are secretly a description of how prime numbers are scattered along the number line. Get the zeros right and you understand the deepest regularity governing the primes — the atoms of arithmetic. Get them wrong, or leave them unknown, and a foundational uncertainty runs underneath vast tracts of modern mathematics, silently qualifying thousands of theorems with the phrase "assuming the Riemann Hypothesis."

svg[The critical strip: zeros found so far all sit on the line Re(s) = 1/2]

The functional equation also explains the "trivial" zeros: at negative even integers, the $\Gamma$ -function factors force $\zeta$ to vanish for reasons that have nothing to do with primes — bookkeeping zeros, essentially, outside the strip that matters. Everything of number-theoretic interest lives in the strip, and everything checked so far lives on its spine.

By the early 20th century, Hardy had proven infinitely many zeros lie exactly on the critical line — an existence result, not a completeness one. Later work by Selberg, and then by Levinson, Conrey, and others, pushed this toward showing a positive proportion of all zeros lie on the line: the current record, from Conrey's 1989 work refined since, shows at least 41% of nontrivial zeros are on the critical line. That number has crept up over decades of hard analytic work but is nowhere near 100%, and there is no known technique that scales to "all."

Where it breaks down

Here is the honest state of the evidence, and where confidence has to yield to uncertainty.

The computational record is spectacular but logically almost beside the point. Since Turing's pioneering hand-and-machine verification in 1953 (he built a method — still used in spirit today — for rigorously certifying that all zeros up to a given height lie on the line, not merely the ones you happened to find), the count has grown through the Riemann Hypothesis verification projects of Brent, van de Lune, te Riele, Odlyzko, Gourdon, and Platt, among others, to more than $10^{13}$ zeros confirmed on the line, with no exception. But a false statement about infinitely many objects can be true for the first googol of them and fail at the googol-and-first. Number theory is littered with cautionary tales: Skewes' number arose because $\pi(x) < \mathrm{Li}(x)$ was believed to hold always, based on all computable evidence, until Littlewood proved the inequality must reverse infinitely often — the first proven crossover lurks at a bound so large it dwarfs any number ever computed directly. The lesson computational number theorists draw is blunt: primes have long memories, and violations can hide arbitrarily far out. Ten trillion zeros is reassuring, not decisive.

More structurally troubling: nobody has a proof strategy that visibly scales. The 41%-and-rising results use delicate but essentially combinatorial/analytic estimates (mollifiers, moment computations) that get harder to push as you demand a higher proportion; there's no known argument that turns "a positive proportion" into "all," any more than knowing most people in a room are under six feet tall tells you the tallest person is under six feet. The zero-free region techniques that gave the Prime Number Theorem its original proof — showing $\zeta(s)\neq 0$ in a thin sliver approaching the line $\mathrm{Re}(s)=1$ — have been improved since 1899, but only by constant factors and logarithmic tweaks (results of Vinogradov, Korobov, and more recent refinements by Ford and others). None come close to the full strength RH demands.

And there is a subtler, more philosophically uncomfortable gap: we don't have a reason, in the sense of a structural mechanism, for why the zeros should be forced onto the line. The nearest thing to a conceptual explanation is a beautiful analogy with physics, discussed below, but an analogy is not a proof, and analogies in mathematics have failed before. The honest state of affairs is that RH is supported by (a) unbroken numerical evidence at enormous scale, (b) a proven positive-density result, (c) an almost-too-good-to-be-coincidence resemblance to spectral theory and random matrix statistics, and (d) the sheer number of consistent, non-contradictory consequences it has across mathematics. None of these constitute proof. Mathematics has been fooled by strong heuristic evidence before.

The contenders

There is no single rival "theory" competing with RH the way, say, competing quantum gravity programs compete with each other — there's one precise statement, and the open question is whether it's true, false, or (in principle, though this seems unlikely given how concretely it's stated) independent of the usual axioms. But there are several distinct research programs aiming at a proof, each with real appeal and real trouble.

The spectral / Hilbert–Pólya program. The most seductive idea, floated independently by David Hilbert and George Pólya early in the 20th century, is that the imaginary parts $\gamma$ of the nontrivial zeros are eigenvalues of some self-adjoint (Hermitian) operator. Self-adjoint operators, by a basic theorem of functional analysis, have real eigenvalues — full stop, no exceptions, no strays. If the zeros $1/2 + i\gamma$ are secretly the spectrum of such an operator, RH would fall out for free, structurally, the way a symmetry forces a conservation law. The trouble: no one has found the operator. Worse, no one has a first-principles derivation of what it should be, though physicists have proposed candidates inspired by chaotic dynamical systems and quantum mechanics on hyperbolic surfaces. It remains a hope organizing a search rather than a technique that has borne fruit.

Random matrix theory and the Montgomery–Odlyzko phenomenon. In 1972, Hugh Montgomery was studying the pair correlation of zeta zeros — how the spacings between them statistically behave — and, over tea, mentioned his formula to the physicist Freeman Dyson. Dyson recognized it instantly: it was exactly the pair correlation function for eigenvalues of random Hermitian matrices from the Gaussian Unitary Ensemble (GUE), objects studied in nuclear physics to model energy levels of heavy atomic nuclei. Odlyzko's massive numerical computations of zeta zeros in the 1980s confirmed the match to stunning precision. This is one of the strangest and most cited cross-pollinations in modern mathematics: the statistics of prime-encoding zeros behave like the statistics of energy levels in quantum chaotic systems. It has become a research program in its own right, linking analytic number theory to random matrix theory and quantum chaos, and it strongly suggests the Hilbert–Pólya vision is on the right track — some operator with GUE-type spectral statistics should exist. But a statistical resemblance, however precise, is not an identification of the operator, and it does not by itself prove any single zero, let alone all of them, sits on the line.

Function field and geometric analogues — the proven cousin. Here the news is genuinely good, and worth telling honestly because it shows what a real proof of an RH-type statement looks like. Riemann's zeta function has analogues for algebraic curves over finite fields, and André Weil proved the Riemann Hypothesis for those in the 1940s, using deep tools of algebraic geometry over finite fields. Decades later, Pierre Deligne proved the vast generalization — RH for varieties of any dimension over finite fields — as the capstone of the Weil conjectures, work recognized with a Fields Medal in 1978. This is a real, complete, celebrated proof of a true structural analogue of RH. It shows the statement can be true and can be proven with the right machinery — algebraic geometry, étale cohomology, the Frobenius endomorphism acting on cohomology groups with eigenvalues controlled in absolute value by the Weil pairing. The catch, and it is a serious one, is that this machinery is intrinsically geometric, built on varieties over finite fields, and nobody has found a way to transport it to the classical integers and the classical zeta function. The analogy is a landmark and a taunt in equal measure: it proves the idea is not absurd while giving no direct route to the original problem.

Explicit analytic attacks and zero-density results. A less glamorous but steadily productive program simply grinds away at estimates: zero-free regions, zero-density theorems (bounding how many zeros could be off the line, even if not proving none are), and effective versions of the explicit formula linking primes to zeros. This tradition — Hardy, Littlewood, Selberg, Levinson, Conrey, Iwaniec, Ford, and others — has produced real, if incremental, progress: better constants, wider verified regions, larger proven proportions on the line. It is unglamorous compared to Hilbert–Pólya's siren song, but it is the only approach with an unbroken track record of actual, verified theorems strictly stronger than what came before, even if none of them close the gap to "all."

Figure

Two visions of why the zeros might be forced onto the line

What would settle it

What would count as decisive? Two clean, opposite roads.

A disproof is conceptually simple to state even if practically extraordinary: locate one nontrivial zero with real part not equal to $1/2$ , and rigorously verify it — either by a certified numerical computation (as Turing-style methods do for the zeros they confirm on the line, one could in principle certify one off the line) or by an analytic argument showing that some infinite family of zeros must eventually stray. Given the depth of numerical search already conducted, an accidental discovery via brute-force computation on known ranges seems unlikely; a disproof, if it comes, would more plausibly emerge from some clever analytic construction showing the zero-free region cannot be widened past a certain point, or from a counterexample built via unexpected algebraic structure. Most working number theorists consider a disproof unlikely, given the accumulated evidence — but "unlikely according to consensus" has been wrong before in mathematics, and the discipline is careful to say so.

A proof is harder to picture concretely because we don't yet have the right vocabulary. Plausible routes discussed by specialists include:

Actually constructing the Hilbert–Pólya operator — finding a genuine self-adjoint operator (perhaps arising from a still-undiscovered dynamical or quantum system) whose spectrum provably matches the zeta zeros. Alain Connes has pursued a program using noncommutative geometry and an operator related to the adele class space, aiming at exactly this; it remains incomplete but is one of the most sustained serious attempts.
Finding a genuine arithmetic analogue of Deligne's cohomological methods that applies to $\mathrm{Spec}(\mathbb{Z})$ itself rather than to varieties over finite fields — this is the ambition behind speculative frameworks like the "field with one element," $\mathbb{F}_1$ , which tries to imagine $\mathrm{Spec}(\mathbb{Z})$ as if it were a curve over some deeper base, so that Weil's methods might one day transplant across.
Radical improvement of explicit analytic estimates — zero-density theorems strong enough, combined with sieve-theoretic or additive-combinatorial inputs, to squeeze the "at least 41%" proven on the line all the way to 100%, though no one currently sees the mechanism that would make such a squeeze terminate rather than merely improve forever.
An entirely unforeseen bridge, of the kind Yitang Zhang's 2013 breakthrough on bounded gaps between primes, or James Maynard's subsequent refinements, provided for a different hard prime-distribution problem — a technique nobody expected, arriving from an adjacent but not identical direction.

As for timing: there is no responsible forecast. RH has now resisted concerted effort for over 165 years, a span that has seen the invention of most of 20th-century mathematics. Some specialists — Andrew Granville among them — have noted that even our confidence in RH's truth rests substantially on aesthetic and structural intuition (it would be strange, in a sense, for the universe of primes to be this well-behaved for numerological luck alone), rather than on a visible path to proof. It is entirely reasonable, and not defeatist, to say we do not know whether a proof is one clever paper away or another century away — or whether it might instead take a wholly new field of mathematics, not yet founded, the way Deligne's proof needed a mathematics that didn't exist in Riemann's day.

Why it matters

The immediate stakes are almost embarrassingly concrete: hundreds of published theorems in analytic number theory are stated conditionally, "assuming RH" or its generalized cousins (the Generalized Riemann Hypothesis, extending the same statement to Dirichlet L-functions and beyond, which underlies conjectural bounds used in computational number theory and cryptography). Proving RH would instantly upgrade this entire conditional literature to unconditional truth — arguably the single largest one-time promotion of "probably true" to "true" available anywhere in mathematics. Disproving it would force a slower, more painful reckoning: identifying exactly which of those theorems survive on other grounds and which quietly depended on an assumption that turned out false.

Beyond bookkeeping, RH matters because it is a referendum on how lawful the primes really are. The primes are defined by the simplest possible rule — no factors — yet their individual placement looks locally erratic, almost like noise. The prime number theorem tames the average behavior. RH would certify that even the fluctuations around that average are as small and well-behaved as a "random" sequence of the same density would produce — nothing hidden, no conspiratorial clumping at some unimaginable scale. It's a statement, in other words, about whether arithmetic's most fundamental building blocks are secretly pattern-free noise dressed as numbers, or whether a deeper regularity, still invisible to us, might eventually surface.

That question — is there a hidden reason things line up as beautifully as they do? — connects RH to the theme running under two of mathematics' other great open frontiers. The eerie resonance between zeta zeros and random matrix eigenvalues from nuclear physics is one more instance of the puzzle explored under the unreasonable effectiveness of mathematics: why should a statistical law discovered to describe uranium nuclei say anything true about prime numbers, unless mathematics is tapping into some structure more unified than the surface-level subjects suggest? And RH's stubbornness against every known proof technique gives it a family resemblance, at the level of difficulty, to the P versus NP problem: both are precisely stated, believed true by most specialists on strong circumstantial grounds, resistant to every current method, and both would reorganize significant parts of mathematics and computation the moment they fell. Neither problem's resistance is fully understood either — we don't just lack a proof, we lack a convincing account of why the problem is so hard, which is its own open question sitting one level above the original one.

There are also live technological stakes. Fast primality tests and factoring-adjacent algorithms in computational number theory (and by extension, cryptographic security analyses that rely on the presumed difficulty of certain number-theoretic problems) sometimes lean on the Generalized Riemann Hypothesis for their proven running-time guarantees. A disproof, or a proof, would not likely break any cryptosystem overnight — RH's role there is more about certified performance bounds than about secrecy itself — but it would reshape which algorithms carry unconditional guarantees versus conditional ones.

The people chasing it

The problem has drawn, and consumed, extraordinary talent for a century and a half.

Origins · 1 minBernhard Riemann and how this was discovered

Bernhard Riemann

1826–1866 · German mathematician whose eight-page 1859 paper "On the Number of Primes Less Than a Given Magnitude" introduced the zeta function's analytic continuation, the functional equation, and the hypothesis itself — almost as an aside, in a sentence he did not attempt to prove

physicist[G. H. Hardy|1877–1947|Proved in 1914 that infinitely many nontrivial zeros lie on the critical line; a famous story holds he once listed proving or disproving RH among things he wanted to accomplish before a dangerous sea crossing, reasoning God would not let him die with the glory of refuting him still unclaimed]

Origins · 1 minAlan Turing and how this was discovered

Alan Turing

1912–1954 · Devised, in 1950s Manchester, some of the first rigorous computer-verification methods for confirming that all zeros up to a given height lie exactly on the line — establishing the template later verification projects would scale to trillions of zeros

physicist[Atle Selberg|1917–2007|Developed the Selberg trace formula and Selberg's sieve, proved a positive proportion of zeros lie on the critical line, and helped forge the deep, still-mysterious link between zeta zeros and spectral theory on hyperbolic surfaces]

Origins · 1 minHugh Montgomery and how this was discovered

Hugh Montgomery

b. 1944 · Discovered the pair-correlation conjecture for zeta zeros in the early 1970s, whose accidental resemblance to random matrix statistics — spotted instantly by Freeman Dyson — opened the entire random-matrix research program

physicist[Pierre Deligne|b. 1944|Proved the Weil conjectures in the 1970s, including the finite-field analogue of the Riemann Hypothesis, using étale cohomology — the clearest existing proof that an RH-type statement can be true and provable, in a cousin setting, with the right machinery]

Origins · 1 minAlain Connes and how this was discovered

Alain Connes

b. 1947 · Has pursued, since the 1990s, a noncommutative-geometry approach aiming to realize the Hilbert–Pólya vision concretely, constructing spectral systems related to the zeros via the adele class space, part of an unfinished but rigorously pursued program

history

1859 — Riemann's memoir introduces $\zeta(s)$ 's continuation and states the hypothesis
1896 — Hadamard and de la Vallée Poussin independently prove the Prime Number Theorem
1914 — Hardy proves infinitely many zeros lie exactly on the critical line
1972 — Montgomery's pair correlation conjecture connects zeta zeros to random matrix theory
1974 — Levinson shows at least a third of zeros lie on the line; later improved past 41% by Conrey

The open frontier

Stand back from the technical apparatus and the shape of the problem is almost unbearably clean: a single, sharply stated claim about where certain complex numbers sit, unresolved despite more raw computational and human-intellectual effort thrown at it than almost any other question in the field. It is the kind of problem where every tool built to attack it has instead illuminated something else first — the Prime Number Theorem fell to a weaker zero-free statement decades before anyone dared hope for the full line; the finite-field analogue fell to machinery invented for entirely different geometric purposes; the statistical fingerprint of the zeros turned out to belong to nuclear physics before anyone connected it back to number theory. RH seems to reward every serious effort with treasure — just never quite the treasure being sought.

That pattern is itself worth sitting with. Mathematics doesn't usually behave this way, offering so much collateral richness while withholding the central prize for so long. It suggests either that the missing idea is genuinely foreign to everything tried so far — waiting in a field not yet invented, the way cohomology theory was waiting, unimagined, in Riemann's century — or that the statement is simply harder than its clean phrasing lets on, demanding a depth of structural understanding about the integers that we are still, after a century and a half, only beginning to acquire.

What remains genuinely open, stated plainly:

Does a self-adjoint (Hilbert–Pólya) operator whose spectrum is the zeta zeros actually exist, and if so, what natural mathematical or physical system produces it?
Can the proportion of zeros provably on the critical line be pushed from its current 41%-and-rising toward 100% by any known method, or does that approach have an inherent ceiling?
Is there a genuine arithmetic analogue of Deligne's cohomological proof of the finite-field Riemann Hypothesis that could apply to the integers themselves?
If RH is true, is it provable within the ordinary tools of arithmetic, or could it be the kind of true-but-unreachable statement that Gödel's incompleteness results show must exist somewhere in any sufficiently rich formal system?
What, if anything, would a single counterexample destroy — which of the thousand conditional theorems collapse, and which survive on independent footing?

Ask the tutor: If the zeros are truly the "spectrum" of some hidden operator, what kind of physical or mathematical system would you go looking for it in — and what would it feel like to actually find it?